In this article, Prof. Soloveitchik attempts to demonstrate external influence in the Halachic thought of the Tosafists. He does this by pointing out that the Halachic basis cited in favor of suicide in the face of conversion is remarkably weak, proving that "the conclusion arrived at by the thinker is so atypical that unless something impinged, consciously or unconsciously,upon his thought he could never have arrived at the conclusion that he did."
His hasagot on the reasoning of the Tosafists is as follows:
1. It's remarkably unusual for the aggadot to be cited as a normative source.
2. The inferences drawn by the Tosafists from aggadic material are systematically and grossly flawed, in that the stories cited are not cases in which a choice was given to "transgress or be killed",etc.
I believe neither of these hasagot are serious for the following reason. The term יהרוג of יעבור ואל יהרוג is somewhat ambiguous (as Soloveitchik points out regarding יעבור at the end of Part 2 of his article) - the words "he should be killed" can refer either to 1. allowing oneself to be killed or, 2. Suicide. The problem facing the Tosafists was as follows: Is suicide also included within the rubric of the term יהרוג or only murder?
To solve what is basically a question of terminology, there is no reason that the aggadot shouldn't be used (somewhat akin to the giluy milsa (See Ritva Kiddushin 2a and Birurei Hamidos V. 2 Chapter 3) which is merely cited as a lexicographical proof as opposed to the Gezeirah Shaveh which is an actual source.) In this case the fact that in a general manner (even if the specifics don't always match) we see in several aggadot that suicide is an option, is enough of a proof that יהרג refers to suicide.