Probability in the eye of the beholder?

In his debate with Daniel Dennett on whether science and religion are compatible, Alvin Plantinga asks the following question:

Let D be the proposition that the variety of the living world has come to be by Darwinian processes, E the relevant biological evidence, G the proposition that evolution is guided, and U the proposition that it is unguided. Then our question is which is greater: P(D/E&G) or P(D/E&U)? (Science and Religion: Are They Compatible?, p.12)

"P(D/E&G)" means "the probability of D, given E and G" and "P(D/E&U)" means "the probability of D, given E and U". Plantinga claims that P(D/E&G) > P(D/E&U). His grounds for this claim are:

1. "Clearly God could have created living things by way of natural selection, causing the right mutations to arise at the right time, preserving the relevant populations from disaster, and the like."
2. "The eye, the mammalian brain, and other organs remain difficult problems for unguided evolution."
3. "[T]he stupefying complexity of the living cell, both prokaryotic and eukaryotic."

1 is his reason for thinking that "P(D/E&G) is perhaps not terribly low", while 2 and 3 are his reasons for thinking that "P(D/E&U) is exceedingly low". "Not terribly low" and "exceedingly low" are both vague expressions, but no doubt everyone would agree that "exceedingly low" is lower than "not terribly low". Indeed, Plantinga thinks it is "orders of magnitude lower" .

It is clear that 2 is just the old argument from complexity and 3 is the new(ish) argument from irreducible complexity. Let us put aside the fact that both 2 and 3 have been adequately addressed by evolutionary biologists, and focus on 1 and the claim that "P(D/E&G) is perhaps not terribly low".

If we looked only at the complexity of some aspects or features of life on earth, we might agree with Plantinga's estimation of P(D/E&G). But complexity is not all there is to D. What about the messiness of evolution and all the evolutionary dead ends? If we look at the latter as well, and if we, like Plantinga, understand the "guided" in G to mean "guided by God", where God is supposed to be omnipotent and omniscient, shouldn't we estimate P(D/E&G) to be lower, perhaps way lower, than Plantinga's "not terribly low"? More simply put, what is the chance of God's having done such a lousy job?

The point I have just made is fairly simple; the interesting question is why Plantinga doesn't see it.