The following seems to me to be a pretty good objection to the view that p is part of S's evidence if p is justifiably believed by S (JBJE).
Let 'p' = Verna is a vixen.
Let 'q' = Verna is a female fox.
Suppose that S believes q on the basis of inductive grounds, ig, and that ig are sufficient for justifiably believing q. I'm not a foxologist, so I don't know what ig would consist of. Describe whatever behavior you think makes it quite likely that Verna is a female fox. S deduces p from q and knows that p and q are logically equivalent. S thus knows that she ought to be equally confident in p and q. If she knows that q is justified for her and knows JBJE, by JBJE, she knows that ~q is inconsistent with her evidence. She thus could deduce that ~p is inconsistent with her evidence. According to JBJE, in acquiring ig, S thereby acquires the sort of evidence such that ~p and ~q are hypotheses inconsistent with S's evidence. Which is crazy, of course, because ~q and ~p are consistent with ig.
Why does this matter? Well, it doesn't. It does show that there's no problem in saying both that evidence consists of truths and that there can be justified beliefs where the propositions believed are not part of the subject's evidence. And that removes a major obstacle to defending the view that evidence consists of truths because it helps to show that it's no consequence of such a view that Gettier cases are impossible.