I don’t see anything wrong with this argument (not having read the full paper.) I myself see epistemology as a conditional probability of our senses being correct: indeed, my personal tendency toward naturalism, (admittedly a peculiar naturalism that would indeed encompass any heretofore-regarded “supernatural entity” as “natural” insomuch as they communicate phenomenally,) is tempered by the knowledge that my senses themselves are derived from the natural world.
Evolution doesn’t really have much to do with it, except that it goes a long way to explain how the senses are so go at gathering data from the natural world: indeed, that is why I weigh the conditional probability of the scientific epistemology so heavily, because it is one of the few epistemologies that achieves actual results.
I haven’t read the article (although I have read similar arguments). If the wiki reconstruction of his argument is correct, then there is an obvious slide in his argument which, in my opinion, blunts its force. He starts out by talking about whether evolution would favor truth-apt cognitive machinery, but then develops his argument by talking about whether evolution would favor true beliefs, rather than merely adaptive beliefs. But evolution doesn’t select for beliefs at all. Our beliefs about the world are not (at least by and large) innate; they are learned through experience by means of our evolutionarily-produced cognitive mechanisms. And so what Plantinga needs to argue is that evolution could produce a cognitive mechanism which systematically produced beliefs which were false, yet somehow adaptive. The evolution of a system which in response to novel and unpredictable input from an ever-changing world produces systematically false yet consistenly adaptive beliefs would be miraculous; the chances are quite small. What would be the easiest and most obvious way for a cognitive system to produce adaptive beliefs? By producing true beliefs, which would accurately guide us in navigating the world and the hazards it contains.