Full opinion text
HAYS, Circuit Judge: On September 6, 1970 Pan American Flight 083, while on a regularly scheduled flight from Brussels to New York, was hijacked in the sky over London about 45 minutes after it had taken off from an intermediate stop in Amsterdam. Two men, Diop and Gueye, acting for the Popular Front for the Liberation of Palestine [the “PFLP”], forced the crew of the aircraft to fly to Beirut, where a demolitions expert and explosives were put on board. The aircraft, a Boeing 747, was then flown to Egypt still under PFLP control. In Cairo, after the passengers were evacuated, the aircraft was totally destroyed. We are asked on this appeal to determine which of the various underwriters that insured the aircraft must bear the cost of the loss. This determination depends on whether the September 6 hijacking was proximately caused by an agency fairly described, for insurance purposes, by any of the exclusions contained in a group of identical all risk aviation policies — policies which, if not for the exclusions, would cover the loss. The district court held that the all risk policies covered the loss. Pan American World Airways, Inc. v. Aetna Casualty and Surety Co., 368 F.Supp. 1098 (S.D.N.Y.1973). I. Insurance Policies in Suit There is no dispute as to the fact of the loss of the 747, as to the amount of the loss, or as to the provisions of the various insurance policies potentially covering the loss. The controversy on this appeal involves the interpretation of those policies. The aircraft in question was covered by “a more or less seamless mosaic” of insurance policies, 368 F.Supp. at 1101, distributing among three classes of insurers, the risk of loss depending on the proximate cause of the damage. Members of the first class of insurers wrote identical aviation all risk policies. These policies indemnified Pan American against “all physical loss of or damage to the aircraft,” except for any loss “due to or resulting from” certain specified exclusions. The all risk policies became effective as a group on November 12, 1969, and covered the aircraft for the following year. They covered damage or loss in any amount up to the full agreed upon value of the 747, to wit, $24,288,-759. Pan American paid a premium of $4,571,635 for this coverage for its entire 747 fleet for the period of January 1,1970 to September 21,1970. This first class of insurers, to be referred to as the “all risk insurers,” included three separate groups. Members of the United States Aviation Insurance Group [the “USAIG”], including Aetna Casualty and Surety Co., participated in the all risk insurance to the extent of one-third of the agreed upon value of the Pan American fleet. The insurance was written on USAIG forms. Members of Lloyd’s underwriting syndicate [the “London all risk insurers”], including David Linton Dann, participated to the extent of one-sixth of the agreed upon value. Members of the Associated Aviation Underwriters [the “AAU”], an American group, participated to the extent of one-half of the agreed upon value by way of reinsurance of the Federal Insurance Co. Parenthetically, it appears that these three groups include all of the underwriters in the world who write aviation all risk insurance for American air carriers. The exclusions in the all risk policies, insofar as they are relevant here, read as follows: “34. LOSS OR DAMAGE NOT COVERED “C. This policy does not cover anything herein to the contrary notwithstanding loss or damage due to or resulting from: 1. capture, seizure, arrest, restraint or detention or the consequences thereof or of any attempt thereat, or any taking of the property insured or damage to or destruction thereof by any Government or governmental authority or agent (whether secret or otherwise) or by any military, naval or usurped power, whether any of the foregoing be done by way of requisition or otherwise and whether in time of peace or war and whether lawful or unlawful (this subdivision 1. shall not apply, however, to any such action by a foreign government or foreign governmental authority follow-the forceful diversion to a foreign country by any person not in lawful possession or custody of such insured aircraft and who is not an agent or representative, secret or otherwise, of any foreign government or governmental authority) [hereinafter “clause 1”]; 2. war, invasion, civil war, revolution, rebellion, insurrection or warlike operations, whether there be a declaration of war or not [hereinafter “clause 2”]; 3. strikes, riots, civil commotion [hereinafter “clause 3”]. The only issues in this appeal involve the interpretation of these exclusions. Effective January 1, 1970, Pan American obtained coverage on the London war risk market for losses caused by the perils excluded by the all risk policies. The terms of this coverage were contained in so-called war risk policies that were issued by Lloyd’s underwriters, including Philip Gaybell Wright, and various other participants in the London war risk market. These insurers will be referred to jointly as the “war risk insurers.” At the time of the loss, September 6, 1970, the language specifying coverage in these policies was precisely identical to the language specifying exclusions in the all risk policies. The upper limit of war risk coverage was $14,226,290.47. Pan American paid a premium of $190,-511 for this coverage for its 747 fleet for the period January 1 through September 12. The London market was and is the only private source of aviation war risk insurance. American underwriters do not write war risk coverage. Thus, Pan American had to turn to the United States government for war risk coverage for the excess over the London market limit. The Secretary of Transportation is authorized by Title XIII of the Federal Aviation Act of 1958, as amended, 49 U.S.C. §§ 1531-1542 (1970), to issue insurance covering the types of risks normally excluded under “free of capture and seizure” clauses [“F.C.&S.” clauses] similar to clauses 1 and 2 of the all risk exclusions. Accordingly, for a premium of $45,000, the government issued a policy covering specified war risks in the amount of $9,763, $709.53 in excess of $14,226,290.47. The total limit of London and government war risk coverage was $24,000,000, or about $288,000 less than the agreed upon value of the aircraft. The government policy covered loss or damage “resulting from” the following perils: “War, invasion, acts of foreign enemies, hostilities (whether war be declared or not), civil war, rebellion, revolution or insurrection, military or usurped power or confiscation and/or nationalization or requisition or destruction by any government or public or local authority or by any independent unit or individual engaged in irregular warfare.” (Emphasis added.) In the district court, the government conceded that the italicized language describes the present loss, unless the loss was due to “strikes, riots, [or] civil commotion.” But the government policy does not cover war losses to the extent that there is double insurance coverage. It disclaims “[a] Any liability or claim for injury, loss, damage or expense covered under any other policy of insurance, including any guaranty or indemnity agreement, in effect for the benefit of the Insured; the Insured warrants that the Insurer shall be free from any such liability or claim.” Thus, the government must pay the 9.8 million dollar excess if the loss was proximately caused by an event described by the clause 1 or the clause 2 exception to all risk coverage. But if the loss was caused either by “strikes, riots, [or] civil commotion” as described in the clause 3 exception, or by any cause not excluded from the all risk insurance, the government is not liable for the excess. Pan American closed the final major gap in its insurance coverage in July, 1970, when for an additional premium of $29,935, the American all risk insurers “wrote back” coverage, that is deleted the exclusion, for the clause 3 risks to the extent of $10,062,393 in excess of $14,226,290. Members of the USAIG and the AAU each underwrote half of this additional coverage. The interest of the insured and the various insurers in the interpretation of the all risk exclusions may be summarized as follows: If the loss was proximately caused by a clause 1 peril (“capture, seizure ... or any taking . by any military ... or usurped power”), or a clause 2 peril (“war . . . civil war, revolution, rebellion, insurrection or warlike operations”), Pan American will recover $24,000,000, approximately $14,200,000 of which will be paid by underwriters in the London war risk market, and approximately $9,800,000 of which will be paid by the United States government. If the loss was proximately caused by one of the risks described in clause 3 of the all risk exclusions (“riots, civil commotion”), Pan American will recover approximately $24,300,000, of which $14,200,000 will be paid by the London war risk market, and approximately $10,000,000 will be paid in two equal shares by members of the USAIG and AAU. If none of the all risk exclusions describes the proximate cause of the loss, Pan American will recover $24,-300,000 from the all risk insurers, one-third from USAIG members, one-sixth from participants in the London all risk market, and one-half from members of the AAU. In any event, Pan American is entitled to receive approximately $5,-000,000 in pre-judgment interest. II. The District Court Proceeding When in response to Pan American’s claims all of the insurers denied coverage, Pan American brought the present diversity action in the Southern District of New York, stating claims in the alternative against the three classes of insurers. The all risk insurers cross-claimed for a declaratory judgment that the war risk policies covered the loss, a cross-claim that the district court properly dismissed as frivolous. 368 F.Supp. at 1142. The all risk insurers took the position in the district court that the destruction of the 747 was covered by the clause 1 exclusion for “damage or destruction . . by any military ... or usurped power”; that it was also covered by each of the clause 2 exclusions except that of “invasion”; and that it was due to “riots” and “civil commotion” as those terms are used in clause 3. See 368 F.Supp. at 1117. The war risk insurers took the position that the loss was not “due to or resulting from” any of the excluded risks. Pan American took the position that the loss was not due to an excluded risk, and alternatively, that if it was due to an excluded risk, it was caused by “riots” or “civil commotion.” The government’s position shifted periodically, but it finally seemed to rest on its claim that the loss was due to Pan American’s barratry. 368 F.Supp. at 1141. After the usual extended procedural preliminaries, the case came to trial before Judge Frankel, sitting without a jury. The trial lasted twenty-four days, during which time the court heard the testimony of thirty witnesses, many of whom were brought to Foley Square from distant countries. Depositions of eleven other witnesses were admitted, along with 439 exhibits. The record on appeal includes 267 documents which fill six file drawers. At the trial, the all risk insurers put in an immense amount of evidence relating to the history of war and political tension in the Middle East, and to the climate of political unrest in Jordan. They undertook to show that the hijacking was caused by PFLP or Palestinian Arab “military ... or usurped power,” by offering evidence that these groups operated as paramilitary quasi-governments in parts of Jordan, independently of King Hussein’s authority. They relied on the same kind of evidence and an asserted PFLP intent to overthrow King Hussein, to establish that the loss of the 747 was caused by an “insurrection” in Jordan. With regard to the applicability of these two exclusions, they claimed that various “Fedayeen,” organizations of Palestinian refugees seeking to return Israel to Arab control, had power over substantial territory from which they excluded government functionaries, and that the Fedayeen engaged in violent clashes with the Jordanian Army, culminating in a 10-day “civil war” following the September 1970 hijacking. The all risk insurers also attempted to bring the loss within the scope of the term “war” and “warlike operations” by showing that the PFLP engaged in “guerilla warfare.” Finally, they sought to connect the Pan American hijacking with other hijackings committed on the same day, arguing that all of the hijackings together constituted a single “civil commotion.” A great deal of the evidence adduced by the all risk insurers was double and triple hearsay properly admitted under the relaxed rules of evidence applicable in bench-tried cases, but of dubious probative value. Much of the all risk testimonial and documentary evidence related to propaganda statements made by the PFLP and other Fedayeen organizations, statements which for the most part were not designed for factual accuracy. The testimony of various individuals, relating as it does to emotionally explosive Middle Eastern events, frequently betrayed the bias of the speaker. There was reliable evidence as to the facts immediately surrounding the hijacking. Otherwise, the evidence consisted largely of hearsay, propaganda, speculation and conjecture. The war risk insurers introduced evidence supporting their view of the Middle Eastern situation and the nature and goals of the PFLP. They also sought with some success to undermine the testimony of witnesses for the all risk insurers. For example, they demanded that the testimony of each witness relate to the PFLP, rather than to the Fedayeen in general, cf. 368 F.Supp. at 1108, or if such relation was impossible, that the testimony be appropriately qualified. They constantly pinned down the all risk witnesses in terms of time and space, providing material for their argument that the hijacking over London was not proximately caused by any of the remote events in the Middle East. The war risk insurers introduced evidence establishing so-called insurance facts. For example, they sought to prove that in May, 1970, a USAIG body circulated a memorandum discussing, among other things, the ambiguity of the all risk exclusions in the context of a political hijacking. At the end of the trial the district court was faced with the task of resolving the extensive conflicts in the evidence. After considering post-trial memoranda submitted by the parties, Judge Frankel held that the loss of the 747 was not proximately caused by any peril excluded from coverage by the all risk policies, and he therefore granted judgment for Pan American against the all risk insurers. The district court made detailed findings as to the goals and structure of the Fedayeen in general and the PFLP in particular. Id. 368 F.Supp. at 1107-1109. It found that the PFLP was a militant Marxist-Leninist-Maoist organization, which received financial support and arms from China and North Korea. The PFLP had approximately 600 to 1200 members, 150 of whom constituted a permanent core. Though the PFLP’s primary enemy was Israel, it also condemned “reactionary” Arab regimes, “universal capitalism” and the United States as its enemies, and it was hostile toward the other, generally more moderate, Fedayeen groups. The district court traced the history of Fedayeen activity against Israel, activity which reached a high point in 1968 when the Palestine Liberation Army participated in the Battle of Ka-rama. Id. at 1108. It detailed the rise and fall of the Fedayeens’ prestige among the members of the Palestinian refugee community. The district court found that the purpose of the “external operations” being carried on by the PFLP at the time of the destruction of the Pan American plane was to bolster the morale of the Palestinians, to aggrandize the PFLP’s position in relation to the other Feda-yeen groups, and to call world attention to the plight of the Palestinian refugees. Id. at 1110. The court concluded that the PFLP was a small, isolated group pursuing its own long term objectives. Id. The district judge paid particular attention to the events in Jordan beginning in February 1970. See id. at 1109, 1124-1126. Upon reviewing the evidence, he concluded that the hijacking was not “due to or resulting from” the violent events of 1970 in Jordan. Proceeding to its legal conclusions, the district court found that under governing New York law an ambiguity in a term of exclusion will be resolved in the manner least favorable to the insurer. Id. at 1118. It held that “war” means a conflict between governments, not political groups like the PFLP. Id. at 1130. It further held that the loss was not due to a PFLP “warlike operation” because that term does not include the inflicting of damage on the civilian property of non-belligerents by political groups far from the site of warfare, particularly when the purpose is propaganda. Id. The district court also held that the loss was not due to the acts of the PFLP as a “military or usurped power.” First, the PFLP did not qualify as such a power since it held territory only at the sufferance of the Jordanian government, the de jure sovereign. Second, even if the PFLP was a “power” in Jordan, it did not act as such when it hijacked a plane over London. Id. at 1129-1130. In addition the court held that the PFLP activity was not part of an “insurrection” since there was insufficient evidence that an insurrection against the Jordanian government was in progress at that time, and even if there was an insurrection, the hijacking in question was not primarily caused by it. Id. at 1123-1129. The loss was not due to “riot” because the hijacking was not accompanied by the sort of uproar or disorder that riot connotes in current usage. Id. at 1132-1136. Finally, the district court held that the phrase “civil commotion” comprehends a local disorder rather than a hijacking occurring in the skies over two continents. Id. at 1136-1139.' In short, the district court held that the all risk insurers failed to meet their burden of proving that the cause of the loss was fairly within the intended scope of any of the exclusions. It found that the ancient marine insurance terms selected by the all risk insurers simply do not describe a violent and senseless intercontinental hijacking carried out by an isolated band of political terrorists. The all risk insurers appeal, assailing the district court’s findings of fact as to the size, power, organization, intent and purpose of the Fedayeen and the PFLP, and as to the lack of connection between the hijacking and the larger Middle Eastern events. They attack the district court’s interpretations of the contract clauses, claiming that these interpretations are not supported by the weight of authority, particularly the authorities founded on the law of England. Pan American and the government cross-appeal. We affirm. III. The Hijacking Although there is considerable controversy surrounding the nature, organization and objectives of the PFLP, the facts immediately connected with the hijacking are essentially undisputed. On September 6, 1970, two men traveling on Senegalese passports bearing the names Diop and Gueye purchased tickets in Amsterdam for Pan American Flight 093. About two hours before that purchase, they had attemped to take El A1 Flight 219, but because El A1 security personnel became suspicious of them, they were not permitted to purchase tickets. Diop and Gueye also aroused Pan American suspicions, but they were allowed to board Flight 093 after being searched. See 368 F.Supp. at 1112-1113. Forty-five minutes after Flight 093 departed from Amsterdam, Diop and Gueye produced handguns and grenades, and took control of the aircraft, ordering the crew to fly to Beirut, Lebanon. They had initially planned to take the aircraft to Dawson’s Field, a rudimentary airstrip in northeastern Jordan that had been “occupied” by the PFLP on the same day, but they abandoned that plan when they were convinced that a 747 could not land at Dawson’s Field. The hijackers explained their purpose in the following words: “p.f.l.p. is speaking “why we take the airoplane? we took the American Airoplane because the government of America helps Israel daily — The government of America gives Israel Fantom airoplanes which attack our camps and burn our village. We — The Group of AKA— which is following for p.f.l.p. know that by warning the people of America for the crimes and murders which is committed always on Palestine and Vitnam — make him feel how his government help the Zionism. “We left our homes, and our lands of 20 years old, Every day the Jews attack us — in our Camps — We think that by our work make you know the truth “We are sorry for what we make of disturbance but you must understand us.” When the aircraft came within radio range of Lebanon, the hijackers demanded that PFLP representatives be brought to the Beirut airport. Radio contact was established between the hijackers and Lebanese PFLP leaders Abu Khaled and Abu Ahmad. After Diop and Gueye threatened to blow up the 747 in mid-air if not permitted to land, Lebanese officials reluctantly granted permission on the condition that it take off again after refueling. Once on the ground, seven or eight PFLP members came aboard and placed explosives in the aircraft. One of these additional members, evidently an explosives expert, stayed on board for the next leg of the flight. The 747 took off for Cairo, still under PFLP command. Again after PFLP threats, Egyptian officials reluctantly gave their permission for it to land. The explosive fuses were lit while the aircraft was still in the air. After landing, the passengers were evacuated in good order. The explosives detonated on schedule, and the 747 was a total loss. On the same date, about two hours before Flight 093 was hijacked, the PFLP successfully hijacked TWA Flight 741, forty-five minutes after it left Frankfurt, and Swissair Flight 100, ten minutes after it left Zurich. An attempt to hijack El A1 Flight 219 was foiled in the air by El A1 security personnel. The TWA and Swissair aircraft were forced to fly to Dawson’s Field, where they were surrounded at a distance by about 200 to 1000 Jordanian troops. Explosives were placed around the aircraft by the hijackers. The PFLP held the passengers as hostages, demanding the release of PFLP members held prisoner in Germany, Great Britain, and Switzerland. On September 12, the airplanes were destroyed. IV. Preliminary Legal Issues The all risk insurers, if they were to prevail, had the burden of proving that the proximate cause of the loss of the 747 was included within one of the terms of exclusion. The all risk insurers’ task is made even more difficult by the rule that exclusions will be given the interpretation which is most beneficial to the insured. The maxim contra proferentem receives added force in this case because the all risk insurers knew that their exclusions were ambiguous in the context of a political hijacking, and they knew of but did not employ exclusionary terms used by other all risk insurers which would have clarified the ambiguity. Finally on this appeal, the all risk insurers have the burden of demonstrating that the trial judge’s findings of fact adverse to them were clearly erroneous. A. Burden of Proof. In the district court the insured had the burden of proving the existence of the all risk policies, and the loss of the covered property. Neither of these elements is disputed. Thus Pan American began the action with a prima facie case for recovery which the all risk insurers could meet only by proving that the cause of the loss came under one of the terms of exclusion. B. Contra Proferentem. The loss in this case is covered by the all risk policies if Pan American or the war risk insurers can formulate a reasonable interpretation of the terms of exclusion to permit coverage. On the other hand, it is not sufficient for the all risk insurers’ case for them to offer a reasonable interpretation under which the loss is excluded; they must demonstrate that an interpretation favoring them is the only reasonable reading of at least one of the relevant terms of exclusion. Sincoff v. Liberty Mutual Fire Insurance Co., 11 N.Y.2d 386, 390, 230 N.Y.S.2d 13, 15, 183 N.E.2d 899, 901 (1962). Sincoff states the well-established New York rule for all types of insurance. See, e. g., Bronx Savings Bank v. Weigandt, 1 N.Y.2d 545, 551, 154 N.Y.S.2d 878, 882-883, 136 N.E.2d 848, 851 (1956); Hartol Products Corp. v. Prudential Insurance Co., 290 N.Y. 44, 49-50, 47 N.E.2d 687, 690-691 (1943); Rickerson v. Hartford Fire Insurance Co., 149 N.Y. 307, 313, 43 N.E. 856, 858 (1896); Shneiderman v. Metropolitan Casualty Co., 14 A.D.2d 284, 289-290, 220 N.Y.S.2d 947, 951-952 (1st Dept 1961). Contra proferentem has special relevance as a rule of construction when an insurer fails to use apt words to exclude a known risk. Cf. National Screen Service Corp. v. United States Fidelity & Guaranty Co., 364 F.2d 275, 278-279 (2d Cir.), cert. denied, 385 U.S. 958, 87 S.Ct. 394, 17 L.Ed.2d 304 (1966). The evidence indicates that the risk of a hijacking was well known to the all risk insurers. Between 1960 and 1970 over 200 commercial aircraft were hijacked, eight of which belonged to Pan American. International hijacking is the subject of the Tokyo Convention of 1963 and the Hague Convention of 1970.’ Hijacking is forbidden by a federal statute, 49 U.S.C. § 1472 (1970). The specific risk which caused the present loss was known to the all risk insurers at least three months before the inception of the Pan American policies. In August, 1969, the PFLP hijacked a Trans World Airlines 707 to Damascus and seriously damaged it through the use of explosives. TWA’s war risk insurers, who had written policies with coverage clauses similar to the present all risk exclusions, denied liability. The present all risk insurers took no steps to clarify their exclusions even after the Damascus loss made it clear that the London market did not consider the PFLP hijackings to be within the terms of USAIG all risk exclusions. Various exclusionary terms in use or being considered for use prior to the present loss would have excluded the loss had they been employed. The war risk insurers have demonstrated that “hijacking,” “act for political or terrorist purposes,” “irregular warfare,” “intentional damage,” “forceful diversion” and “theft” enjoyed some currency. For example, in 1969 underwriters in the London all risk aviation market began to consider various alternative wordings for a hijacking exclusion. In November, 1969, an organization of such underwriters adopted the text of an exclusion called “AV-48,” which excluded the following: “(f) Unlawful seizure or wrongful exercise of control of the Aircraft or crew in flight (including any attempt at such seizure or control) made by any person or persons on board the Aircraft acting without the consent of the Insured.” The text of AV-48 was known to the USAIG by June, 1969. Had AV-48 been employed by the present all risk insurers, they might well have avoided liability for the loss of the 747. On September 12, 1969, the London all risk market revised AV-48 to make it even more specific. The revised exclusion, AV-48A, includes the following language: “This Policy does not cover claims directly or indirectly occasioned by, happening through or in consequences of:— “(d) Any act of one or more persons, whether or not agents of a sovereign Power, for political or terrorist purposes and whether the loss or damage resulting therefrom is accidental or intentional. “(e) Any malicious act or act of sabotage. “(g) Hi-jacking or any unlawful seizure or wrongful exercise of control of the Aircraft or crew in flight. Any of these three AV-48A clauses, if employed by the appellant all risk insurers, might well have excluded the present loss. Not only does it appear from the record that various clauses which would have excluded the present loss were in common use, but it appears that the General Policy Committee of the USAIG, which supplied the forms for the present all risk insurance, realized by May, 1970, that “[c]urrent war risk exclusions do not appear to be effective against intentional damage such as might be caused by hijackings, by bombs placed in aircraft by political activists, by riotous acts, etc.” See 368 F.Supp. at 1119. When the all risk insurers failed to exclude “political risks in words descriptive of today’s world events,” id., they acted at their own peril. The all risk insurers contend that the insured, Pan American, should not have the benefit of contra proferentem because the exclusions in the all risk policies are not legally ambiguous. Citing cases like State Farm Mutual Automobile Insurance Co. v. Xaphes, 384 F.2d 640 (2d Cir. 1967), and Order of United Commercial Travelers v. Knorr, 112 F.2d 679 (10th Cir. 1940), they argue that the principle of construing ambiguity against the speaker does not apply to insurance terms that have “already been judicially defined.” We find this argument completely unpersuasive on the facts of this case. The all risk insurers overstate when they assert that their exclusions have been subjected to extensive judicial interpretation. Terms such as “insurrection,” “rebellion” and “civil commotion” have received little of the clear judicial construction which' the courts in Xaphes and Knorr found in regard to the terms in those cases. See 384 F.2d at 641-642; 112 F.2d at 682. It is not irrelevant that the various counsel in this case have been able to infer radically different versions of the “established judicial meanings” of the exclusions. The plausibility of several of these intei’pretations is convincing evidence of the ambiguity of the exclusions, and a compelling reason for applying contra proferentem against the all risk insurers. The all risk insurers claim that contra proferentem is not applied where “the dispute is in reality between groups of insurers.” While their statement of principle may accurately represent the law in some jurisdictions, see, e: g., Boston Insurance Co. v. Fawcett, 357 Mass. 535, 538, 258 N.E.2d 771, 776 (1970), it does not state New York law. In London Assurance Corp. v. Thompson, 170 N.Y. 94, 62 N.E. 1066 (1902), the court resolved a dispute between an insurer and a reinsurer by construing the ambiguity of a dubious coverage clause against the author of the clause, the insurer, relying on the maxim of contra proferentem. In any event, the present case is no mere dispute between insurers. The all risk insurers’ frivolous “cross-claim” against the war risk insurers did not convert this action to a “dispute between insurers.” It did not extinguish Pan American’s substantial interest in the insurance policies. If loss was caused by a clause 1 or a clause 2 exclusion, Pan American will suffer serious financial consequences. As the district court accurately observed, 368 F.Supp. at 1122 n. 27, Pan American faces a $288,000 gap in coverage, which certainly creates a substantial interest. Furthermore, if the loss was caused by a clause 1 or clause 2 agency, there is serious doubt as to whether Pan American can collect any of the government excess war insurance, and as to whether it can collect prejudgment interest from the government. The all risk insurers argue, at least insofar as they claim that the loss was proximately caused by “war,” that although contra proferentem is used in construing life insurance policies, it is not employed in construing property policies. They base this distinction on the apparently fortuitous fact that none of the property cases which construes “war” specifically relies on contra pro-ferentem, probably because of a lack of ambiguity under the facts of those cases which for the most part arose during periods of declared war. The all risk insurers also press upon us the novel argument that contra proferentem is applicable to life policies only because there is no alternative source of war life coverage, while it is not applicable in property policies because alternative war risk coverage is available, presumably on the London market. But the existence of alternative sources of insurance cannot affect the scope of all risk coverage. See Equitable Fire & Marine Insurance Co. v. Allied Steel Construction Co., 421 F.2d 512, 514 (10th Cir. 1970). It is no help to the all risk case that Pan American obtained alternate coverage, particularly as to the value of the 747 in excess of 14 million dollars, for which there was no private alternative source of insurance. For the excess war coverage, Pan American was forced to turn to the government, historically also the source of war life insurance coverage. See, e. g., Reinold v. United States, 167 F.2d 556 (2d Cir.), cert. denied, 335 U.S. 824, 69 S.Ct. 48, 93 L.Ed. 378 (1948); Dennchy v. United States, 15 F.2d 196 (S.D.N.Y.1926). The distinction between property and life coverage has no basis in the decided cases or in common sense. The all risk insurers argue that construing the ambiguity of their exclusions against them would not serve the rationale of contra proferentem, because, they claim, the all risk policies at suit are not adhesion contracts since there is no disparity between Pan American’s and the all risk insurers’ bargaining power. However, they offer no authority for the dubious proposition that contra proferentem derives only from the policy favoring victims of adhesion contracts. In any event, the district court correctly found that in the relevant sense the present all risk policies are adhesion contracts. See 368 F.Supp. at 1121-1122. The all risk insurers are in effect all of the underwriters in the world who write such insurance. Pan American had no place else to turn for all risk coverage. The types of risks to be excluded by the all risk policies may have been negotiable to a certain limited extent, but the words describing these risks were not open to negotiation. These words were offered to the insured on a take-it-or-leave-it basis. Pan Amer-can was in no better bargaining position than any other insured. It could negotiate the scope of its coverage, paying a larger or smaller premium according to whether the scope of coverage was more or less extensive, but it had no significant control of the terms defining coverage. Finally, the all risk insurers seek to avoid the special force which contra pro-ferentem is given in cases where a known risk is not excluded by an apt term, by arguing that they had business reasons for not resorting to more specific exclusions. They argue that the risks of owning an aircraft are divided into three categories: “operational” risks, which are normally undertaken by all risk carriers; “war” risks of the type described in clauses 1 and 2 and in the standard marine F. C. & S. clause, and which are normally undertaken by war risk carriers; and “gray area” risks, risks in between operational risks and war risks, which may be undertaken by either all risk or war risk carriers. They claim that “hijacking” is a gray area risk. They argue that they contemplated that the risk of hijacking would be divided into “warlike” hijacking, which risk was to be undertaken by the war risk insurers, and “non-warlike” hijacking, which risk was to be undertaken by the all risk insurers. It was necessary to so divide the risk, they argue, because domestic airlines desired and needed “non-warlike” hijacking coverage, while American insurers had insufficient experience and premium volume to insure a carrier with foreign routes against “warlike” hijacking. They argue, and one of their experts testified, that if “political risk” exclusions had been incorporated in the USAIG forms, American insurers would have lost premiums to the London market. They claim that if they had employed AV-48 in its entirety, Pan American would have faced a gap in coverage, and if they had used that exclusion in part, reinsurance would have been impossible. The district court found as a matter of fact that these business reasons are not to be credited, 368 F.Supp. at 1122, a finding that is not clearly erroneous. The all risk “business reasons” sound like an attempt to extract the premium for insuring against hijacking, while not covering the type of hijacking that must have been of greatest concern to the international air carrier. There is nothing in the exclusionary language chosen by the all risk insurers to suggest a distinction between “warlike” and “non-warlike” hijackings; these dubious phrases appear nowhere in the policies in suit. Even if the all risk insurers intended to distinguish between “warlike” and “non-warlike” hijackings at the time the policies were written, their subjective or hidden intent is not competent to determine the meaning of the words they employed. See Rickerson v. Hartford Fire Insurance Co., 149 N.Y. 307, 43 N.E. 856 (1896); Wilson Sullivan Co. v. International Paper Makers Realty Corp., 307 N.Y. 20, 25, 119 N.E.2d 573, 575 (1954); Hotchkiss v. National City Bank, 200 F. 287, 293 (S.D.N.Y.1911), aff’d, 201 F. 664 (2d Cir. 1912), aff’d, 231 U.S. 50, 34 S.Ct. 20, 58 L.Ed. 115 (1913). From this discussion it should be evident that if the district court erred in its application of contra proferentem, it erred in the direction of giving it too little weight. It found that this “ancient canon” is not a “decisive concern.” 368 F.Supp. at 1118. But the maxim defines the scope of coverage as much as if it were a clause in the all risk policies. It is part of the understanding of the parties. The experienced all risk insurers should have expected .the exclusions drafted by them to be construed narrowly against them, and should have calculated their premiums accordingly. C. Standard of Review. This case comes to us with the district court having resolved all factual issues against the all risk insurers. Rule 52(a) of the Federal Rules of Civil Procedure provides that “[f]indings of fact shall not be set aside unless clearly erroneous, and due regard shall be given to the opportunity of the trial court to judge of the credibility of the witnesses.” Thus, on this appeal, the all risk insurers have the burden of showing that the district court was clearly in error in its findings leading to the conclusion that the loss was not excluded by any term of exclusion. The clearly erroneous standard applies not only to the district court’s characterization of the events surrounding the September 6 hijacking, but also, in some rather uncertain manner, to the district court’s findings as to the meaning of the terms of exclusion. The district court was obliged, as are we, to construe the exclusions as the parties would reasonably have expected them to be construed. 368 F.Supp. at 1139; see Bird v. St. Paul Fire and Marine Insurance Co., 224 N.Y. 47, 51, 120 N.E. 86, 87 (1918); Harris v. Allstate Insurance Co., 309 N.Y. 72, 75-76, 127 N.E.2d 816, 817-818 (1955); Ore & Chemical Corp. v. Eagle Star Insurance Co., 489 F.2d 455, 456-457 (2d Cir. 1973). The question of the parties’ reasonable expectation is essentially a question of fact. To the extent that the district court’s findings as to the scope of the exclusions depends on the intent of the parties, they may not be upset unless clearly erroneous. The legal authorities cited by the parties construing the terms at issue are relevant mainly as determinates of the parties’ expectations. The clearly erroneous rule has an implication that goes beyond the appellate process. When the all risk insurers wrote the present policies, they took the law as they found it. Premiums were set and coverages were purchased in reliance on rules of construction and procedure. The all risk insurers effectively insured against the “peril” that a fact finder would erroneously but not clearly erroneously determine that a loss was not caused by an excluded agency. Nevertheless, the all risk insurers now claim that at least two of the district court’s findings should be reviewed by this court de novo: the finding, apropos “insurrection,” that the Fedayeen and PFLP did not intend to overthrow King Hussein at any relevant time; and the finding, apropos “warlike operations,” that the PFLP hijacked the 747 primarily for propaganda purposes. The all risk insurers argue that the evidence as to these two issues consists of documentary and “third party reports” of PFLP statements. These statements, the all risk insurers argue, may be evaluated by this court de novo because a decision as to their effect does not turn on the credibility of any witness. In support of this position they site United States ex rel. Lasky v. LaVallee, 472 F.2d 960, 963 (2d Cir. 1973) and Orvis v. Higgins, 180 F.2d 537, 539-540 (2d Cir.), cert. denied, 340 U.S. 810, 71 S.Ct. 37, 95 L.Ed. 595 (1950). The all risk argument goes too far. In the first place, Lasky was a case in which a state prisoner was seeking federal habeas corpus relief and the state court record came cold to both the district court and the court of appeals. 472 F.2d at 963-964. Orvis stands for the proposition that a record consisting only of pleadings, depositions, and affidavits may, at the reviewing court’s discretion, be reviewed de novo. It does not control a case like the present one, in which a great deal of the evidence consisted of third party testimony as to the statements and intentions of others. An example of the evidence that the PFLP intended to overthrow King Hussein illustrates the need to accord respect to the district court’s fact findings. General Abdul Razzak El-Yahya, the Commander of the PLA during 1970, testified that at a meeting with PFLP “leadership” at Dawson’s Field 5 days after the September 6 loss, he heard a PFLP representative state that the object of the hijacking was “to explode a revolution in the entire area.” The weight due this testimony depends on demeanor evidence and the accuracy of General El-Yahya’s observational power and memory, his opportunity for knowing the facts, and his apparent bias, prejudice or interest. This testimony is completely unlike the documents considered in Orvis. There is no basis on this record for not following the clear mandate of Rule 52. The district court heard six weeks of testimony. Its opinion reflects a detailed study of and an intimate familiarity with the record. The all risk witnesses were subjected to extensive cross-examination directed at their credibility. There is no warrant on this record for applying any standard of review other than the clearly erroneous test. V. All Risk Exclusions A. The All Risk Position. The all risk insurers rely on all of the following words of exclusion in the all risk policies: “This policy does not cover anything herein to the contrary notwithstanding loss or damage due to or resulting from: “1. capture, seizure ... or any taking of the property insured or damage to or destruction thereof ... by any military or usurped power, whether any of the foregoing be done by way of requisition or otherwise and whether in time of peace or war and whether lawful or unlawful . . . ; “2. war, . . . civil war, revolution, rebellion, insurrection or warlike operations, whether there be a declaration of war or not; “3. . . . riots, civil commotion.” The all risk position is that the terms employed define uninterrupted overlapping areas of exclusion on a continuum of violence. They claim that in terms of approximately increasing scale and organization of violence, “riot,” “civil commotion,” “insurrection,” “military or usurped power,” “rebellion,” “revolution,” “civil war,” “warlike operations,” and “war” exhaust the possibilities, and that the cause of the loss must be described by at least one of the terms. However, each of the exclusionary terms has dimensions besides the level of violence. For example, for there to be a “riot” three or more actors must gather in the same place; for there to be an “insurrection” there must be an intent to overthrow a lawfully constituted regime; for there to be a “war” a sovereign or quasi-sovereign must engage in hostilities. The doctrine of contra proferentem shrinks the all risk “overlapping areas” to mere points on a line of violence. The lacunae between these points include the vast number of nameless causes that are not precisely described by the terms actually employed. The fact that the all risk insurers have chosen to rely on nearly all of the terms of these three exclusions has affected their cause adversely. The district court correctly observed that we can infer from their reliance on so large a number of exclusions that the all risk insurers recognize that each of the exclusions is ambiguous or has only uncertain application to the facts. 368 F. Supp. at 1117; see Sincoff v. Liberty Mutual Fire Insurance Co., 11 N.Y.2d 386, 390, 230 N.Y.S.2d 13, 16, 183 N.E.2d 899 (1962); Silverstein v. Commercial Casualty Insurance Co., 237 N.Y. 391, 393, 143 N.E. 231, 232 (1924). The all risk insurers’ shotgun approach belies its claim that these terms have certain fixed meanings. B. Proximate Cause. The all risk policies exclude “loss or damage due to or resulting from” the various enumerated perils, a phrase that clearly refers to the proximate cause of the loss. Remote causes of causes are not relevant to the characterization of an insurance loss. In the context of this commercial litigation, the causation inquiry stops at the efficient physical cause of the loss; it does not trace events back to their metaphysical beginnings. The words “due to or resulting from” limit the inquiry to the facts immediately surrounding the loss. Standard Oil Co. v. United States, 340 U.S. 54, 58, 71 S.Ct. 135, 95 L.Ed 68 (1950); Airlift International, Inc. v. United States, 335 F.Supp. 442, 449 (S.D.Fla.1971), aff’d, 460 F.2d 1065 (5th Cir. 1972) (mem.). Thus, in Queen Insurance Co. v. Globe & Rutgers Fire Insurance Co., 263 U.S. 487, 492, 44 S.Ct. 175, 176, 68 L.Ed 402 (1924), Mr. Justice Holmes wrote: “[T]he common understanding is that in construing these policies we are not to take broad views but generally are to stop our inquiries with the cause nearest to the loss. This is a settled rule of construction, and if it is understood, does not deserve much criticism, since theoretically at least, the parties can shape their contract as they like.” New York courts give especially limited scope to the causation inquiry. The leading case is Bird v. St. Paul Fire & Marine Insurance Co., 224 N.Y. 47, 120 N.E. 86 (1918) (Cardozo, J.). In Bird, the insured vessel was damaged by a concussion caused by an explosion in a freight yard about a thousand yards from the vessel. The explosion came about when a fire set off a stock of explosives. The insured sued on an insurance policy covering losses caused by “fire.” The Court of Appeals held that the loss was not caused by fire. It ascertained that the scope of causation relevant to the insurance nature of a loss is largely a question of fact depending on the reasonable expectations of businessmen: “The question is not what men ought to think of as a cause. The question is what they do think of as a cause. We must put ourselves in the place of the average owner whose boat or building is damaged by the concussion of a distant explosion, let us say a mile away. Some glassware in his pantry is thrown down and broken. It would probably never occur to him that within the meaning of his policy of insurance, he had suffered loss by fire. A philosopher or a lawyer might persuade him that he had, but he would not believe it until they told him. He would expect indemnity, of course, if the fire reached the thing insured. He would expect indemnity, very likely, if the fire was near at hand, if his boat or his building was within the danger zone of ordinary experience, if damage of some sort, whether from ignition or from the indirect consquences of fire, might fairly be said to be within the range of normal apprehension.” Id. at 52, 120 N. E. at 87. Britain S. S. Co. v. The King, [1919] 2 K.B. 670 (C.A.), aff’d, [1921] 1 A.C. 99 (1920), illustrates how this principle has been applied by the English courts in the context of war-related losses. While on a voyage from England to Alexandria in the company and under the orders of a British escort, the Matiana went aground and was lost, because the convoy had taken a more northerly route than usual to avoid German submarines. The Court of Appeals held that the loss was due to a marine peril, running aground, rather than to a “warlike operation.” It held that the warlike activity of the escorts did not proximately cause the loss: The Crown’s naval authorities ordered the ship to take a general course fraught with maritime perils, but they did not actually order the Matiana aground. [1919] 2 K.B. at 699-700 (per Atkins, L. J.). The House of Lords agreed, at 1121-22 (.per Atkinson, L. J.). [1921] 1 A.C. Decisions in a variety of other jurisdictions follow the same approach. When the Linwood went aground as a result of the Confederates putting out the Hatteras light, the loss was a “consequence” of a marine peril, rather than “hostilities.” Ionides v. Universal Marine Insurance Co., 143 Eng.Rep. 445, 456 (C.P.1863) (per Erle, C. J.). When the John Worthington collided with a minesweeper that was clearing the channel approaches to New York harbor in 1942, the loss was due to the collision, a marine peril, rather than the warlike reason for the minesweeper’s presence in the harbor. Standard Oil v. United States, 340 U.S. 54, 71 S.Ct. 135, 95 L.Ed. 68 (1950). When the Napoli was lost in a head-on collision with another ship because it was sailing without running lights under British order, the loss was due to a marine peril, the collision. Queen Insurance Co. v. Globe & Rutgers Fire Insurance Co., 282 F. 976 (2d Cir. 1922), aff’d, 263 U.S. 487, 44 S.Ct. 175, 68 L.Ed. 402 (1924). When an insured aircraft was lost over Vietnam in a collision with a military aircraft, the loss was due to an aviation peril, notwithstanding that the two aircraft were flying over Vietnam only because there was a war. Airlift International, Inc. v. United States, 335 F.Supp. 442, 449 (S.D.Fla.1971), aff’d, 460 F.2d 1065 (5th Cir. 1972) (mem.). These cases establish a mechanical test of proximate causation for insurance cases, a test that looks only to the “causes nearest to the loss.” Queen Insurance Co. v. Globe & Rutgers Fire Insurance Co., supra at 492, 44 S.Ct. 175. This rule is adumbrated by the maxim contra proferentem: if the insurer desires to have more remote causes determine the scope of exclusion, he may draft language to effectuate that desire. Id.; Fleeney & Meyers v. Empire State Insurance Co., 228 F.2d 770, 771 (10th Cir. 1955). In the present case, events drawn from the general history of unrest in the Middle East did not proximately cause the destruction of the 747. Of course, in some attenuated “cause of causes” sense, the loss may have resulted from the Feda-yeen or PFLP pattern of military operations against Israel, from the domestic unrest in Jordan, or from the most recent of the three wars which prior to 1970 had convulsed the Middle East. But for insurance purposes, the mechanical cause of the present loss was two men, who by force of arms, diverted Flight 093 from its intended destination. In light of the extensive references in the record to the activities of the Fedayeen and the Palestine Liberation Organization [“PLO”], it is important to bear in mind that the loss of the 747 was not proximately caused by the PLO, the Fedayeen, or A1 Fatah. Evidence elicited by the all risk insurers concerning the activities of the PLO and other less homogeneous groups has no bearing on the causes of the present loss. The PFLP was a small political force that most often acted independently from other Palestinian entities. The district court found, with ample support in the record, that “hijackings and other so called ‘external operations’ . . . were unique tactics of PFLP terrorism, almost uniformly opposed, ... by the other far more numerous fedayeen groups,” 368 F.Supp. at 1110. The PFLP boycotted the PLO during 1969, and sent only one representative, an observer, to the May 30, 1970, Seventh National Council Session of the PLO. The PFLP refused to join the Palestine Armed Struggle Command or the United Command, successive military branches of the PLO. The all risk insurers concede that there were vast philosophical differences between the PFLP, which fought “world imperialism,” and the other more moderate Fedayeen groups, which sought only to destroy Israel. A PFLP propaganda statement of September 13, 1970, stated that the PFLP did not act on behalf of the PLO when it hijacked the 747 and the various other aircraft to Dawson’s Field. Other major Fedayeen groups uniformly condemned hijacking as a tactic. As a result of the September 6 hijackings, the Central Committee of the PLO suspended the PFLP from membership. The all risk insurers’ only argument linking the Fedayeen as a group to the present loss is their assertion that when it committed the September 6 hijacking, the PLFP hoped to force the other Fed-ayeen to follow them. But it misses the point in cases like Bird and Ionides, and it puts the cart before the horse, to attribute the cause of a loss to the organizations which the PFLP hoped to influence by taking and destroying the 747. Aside from the above considerations, all of the parties recognize that when a peril results in the owner’s losing control over insured property, any subsequent damage to or loss of the property is attributable to the peril causing the loss of control. In other words, the proximate cause of a loss- resulting from a taking followed by destruction is determined by the nature of the taking. While the events immediately preceding the taking may proximately cause a loss, the events following a taking may not. The insurance authorities uniformly support this rule, and supply examples of its application. When the Llama was taken at sea by the British government during the First World War and subsequently run aground, the loss was due to a “taking at sea,” rather than a marine peril. Standard Oil Co. v. United States (The Llama), 267 U.S. 76, 45 S.Ct. 211, 69 L.Ed. 519 (1925). When the schooner Yankee was seized by Granadian authorities and subsequently deteriorated, the loss was caused by the seizure. Magoun v. New England Marine Insurance Co., 16 Fed.Cas. p. 483 (No. 8,961) (C.C.Mass.1840) (Storey, C. J.). When the Romulus, a neutral ship, was seized by the Japanese during the Russo-Japanese war and was subsequently lost due to a peril at sea, the House of Lords held that the owner could not recover on a marine policy with the usual F. C. & S. clause. Andersen v. Marten, [1908] A.C. 334, 338. See also General Exchange Insurance Corp. v. Kinney, 279 Ky. 76, 129 S.W.2d 1014 (1939); Dole v. New England Mutual Marine Insurance Co., 7 Fed.Cas. p. 837 (No. 3,966) (C.C.Mass.1864). Cf. J. & H. Schieffelin v. New York Insurance Co., 9 Johns. 21 (N.Y.1812); Greene v. Pacific Mutual Insurance Co., 91 Mass. (9 Allen) 217 (1864). The principle that the taking characterizes the loss has been applied by at least one court to an aircraft hijacking. In Sunny South Aircraft Service, Inc. v. American Fire & Casualty Co., 140 So.2d 78 (Fla.Dist.Ct.App.1962), aff’d, 151 So.2d 276 (Fla.1973), the insured aircraft was covered for theft excluding losses “due to war . . . rebellion or revolution.” The airplane was hijacked in the United States and taken to Cuba where it was damaged by a Cuban military plane. The court found that the loss was proximately caused by theft rather than by warlike activity, and accordingly held that the loss was not excluded. If events following a hijacking were permitted to control the insurance nature of the loss, the outcome in any case would vary according to the whim of the hijacker. In the present case the 747 might well have been destroyed in the air over London by two hijackers, rather than in Cairo by a larger group. The parties cannot have intended that the caprice of the hijackers would control the insurance consequences of the loss. It is not relevant in this case that after the aircraft was hijacked by two actors, a third came aboard, or that extensive civil disorders broke out in Jordan nine days after the hijacking. YI. Clause 1 Exclusion The all risk insurers claim that the destruction of the 747 was “due to or resulting from” unrest in Jordan of a type which is fairly described by the first exclusion: “1. capture, seizure, arrest, restraint or detention or the consequences thereof or of any attempt thereat, or any taking of the property insured or damage to or destruction thereof by any Government or governmental authority or agent (whether secret or otherwise) or by any military, naval or usurped power, whether any of the foregoing be done by way of requisition or otherwise and whether in time of peace or war and whether lawful or unlawful (this subdivision 1. shall not apply, however, to any such action by a foreign government or foreign governmental authority following the forceful diversion to a foreign country by any person not in lawful possession or custody of such insured aircraft and who is not an agent or representative, secret or otherwise, of any foreign government or governmental authority)” (Emphasis added). At the outset there is some controversy as to the implication of the second parenthetical of clause 1, the parenthetical beginning “this subdivision . . . .” The all risk insurers claim that the deletion from the exclusion of acts following a “forceful diversion” implies that “forceful diversion” is a risk excluded by clause 1, except as to the particular case described by the parenthetical. The war risk insurers argue that because “forceful diversion” appears in an exception to an exclusion, we must infer that the all risk policies cover “forceful diversion.” Both of these arguments miss the point of the second parenthetical. The all risk insurers represent that it was added to make it clear that the policy covers hijackings of aircraft to Cuba. It was intended and added, no doubt, out of an abundance of caution, to specify the outcome in a specific case. The parenthetical does not apply to the present facts, because there was no “action by a foreign government.” The parenthetical is relevant only in that it shows that the all risk insurers were quite capable of resolving known ambiguities in concrete terms descriptive of today’s events. The all risk insurers’ principal argument under clause 1 is that the loss of the Pan American .747 resulted from its destruction by a “military . . . or usurped power” in Jordan. They claim that as a matter of law military or usurped power embraces “an organized force defying the general enforcement of the laws by force of arms,” and that this definition applies to PFLP and, more certainly, to Fedayeen activity in Jordan. Pan American argues that to be a military or usurped power, a force must control a substantial territory with trappings of state sufficient to constitute it a “de facto government.” The opposing formulations were considered by the district court, but it did not choose between them. It found that the PFLP “occupied” ground in Jordan at the sufferance of the Jordanian government, and held only that such occupation “is surely insufficient” to constitute a military or usurped power. Accordingly it held that the loss was not excluded by clause 1. 368 F.Supp. at 1129-1130. We hold that in order to constitute a military or usurped power the power must be at least that of a de facto government. On the facts of this case, the PFLP was not a de facto govern