A couple of things. Williamson notes that on the account of evidence Goldman seems to prefer:
it is not even clear that one's evidence must be consistent, let alone true ... But there are grave difficulties in making sense of evidential probabilities on inconsistent evidence, since conditional probabilities are usually taken to be undefined when conditioned on something inconsistent (310).
Goldman had taken issue with the claim that evidence can only consist of true propositions and suggested that instead we think of a subject's evidence as that which the subject justifiably believes non-inferentially. Given some standard assumptions about justified beliefs, a subject could have inconsistent justified beliefs. That's the root of the problem, but the problem is easily avoided if we opt for some factive view of evidence since the facts tend to be consistent.
Williamson does consider a variant on Goldman's view, one on which one's evidence consists of those true propositions that one is non-inferentially justified in believing. He says that the problem with this view is that, "such a view is a rather unnatural hybrid; the truth condition is an ad hoc afterthought, not an organic consequence" (311). Goldman could say that it is better to have an ad hoc view than a counterexampled one, but as I rather like the view that Williamson is criticizing let me say why I think it has organicity going for it.
What is a piece of evidence? It's a fact. It's a fact that stands in various relations, but it's a fact at the very least. What does it take for some piece of evidence to belong to you? It belongs to you when it is properly yours, properly treated as yours, etc... What does it take for evidence to properly belong to you? Justification. That's what justification is all about. A belief is justified just when it is proper to treat what is believed as a reason. A view that tells us what evidence you have is going to tell us two things, what a piece of evidence is and what it takes to possess it. Because it tells us these two things, an account that tells us both of these things at once might seem rather unnatural. But, why should that matter? An account of what it is to have a puppy will involve some account of what distinguishes puppies from things that aren't puppies and what is distinctive of the having relation. Any account that didn't look like an unnatural mix of property rights and zoology wouldn't be the right account.