In History of Sexuality 1, Foucault grants the title of "science" to the biology of reproduction (54E / 73F and 154E / 204F ["les sciences biologiques de la reproduction"]) as opposed to the savoir de la sexualité humaine (204F / 154E). But this doesn't contradict Foucault's claim that "sex" is produced by the modern sexuality dispositif, for "sex" is an "artificial unity" grouping together "anatomical elements [and] biological functions" (the connaissance of which had passed the threshold of scientificity [cf Gutting 1989, 252-256]) with all sorts of other things, such as "conducts, sensations, pleasures."
Now the key move of the sexuality dispositif in constituting "sex" as a natural constant is to then give that unity a causal power (154E / 204F). And while we can note that the sexuality savoir did get some "quasi-scientificity" from its "proximity" with the bona fide science of reproduction, we also have to note that the undoubted scientific contents of the biology of reproduction could be recruited to "serve as the principle of normality for human sexuality" (154-155E / 204F).
Using the notion of "threshold of scientificity" Gutting 1989 lays out (252-256), it seems that Foucault would not object to a (Bachelardian / Canguilhemian) history of science that would say "now we know that Aristotle is wrong about reproduction, and that it was really sperm and egg and not the objects postulated by his theory that were at work." But saying this is not an archaeological analysis, so that Foucault could still say that Aristotle produced true statements relative to his episteme. In other words, to be a little paradoxical about it, he is wrong but was true. Wrong by the standards of contemporary science, but true by the standards of archaeology. (Presumably, had Foucault ever done an archeology of biological savoir in Aristotle's time.) [NB: slight rewording of this graf -- inserting the tenses in "is wrong but was true" -- done on Sun 1 April 2:30 pm CDT.]