If somebody believes it's morally wrong to φ, it may still be right for them to φ. People can be mistaken about what it's morally wrong to do. We have a great example in the Huck Finn case, where the title character firmly believes that he is wrong to not send his friend Jim back to slavery, and does the right thing even when he believes it's wrong. There are certainly many more prosaic cases.
If somebody believes it's irrational to φ, I think (contra Michael Smith) that it may still be rational for them to φ. They may accept a bad theory of practical rationality. If Humeans are right about practical rationality, Kantians may falsely believe that they're acting wrongly in a variety of cases.
Now it's looking to me like justification for belief will have to go the same way. If I believe that I am unjustified in believing that p (or even that I am justified in believing not-p), I may still be justified in believing that p. I could merely be in the grips of a bad theory of epistemic justification. Maybe I just talked with a very convincing and charismatic external-world skeptic. That won't make me generally unjustified in my beliefs about the world. (If it does, the skeptic is in a better position than we usually take him to be!)
I think this conclusion would probably be a little more surprising to people, because it sounds like some kind of Moore's paradox variant. And since the mental states in question here are both beliefs (a belief about justification, and a belief that p) we might think that they're supposed to interact with each other in the mind of a rational agent, with one causing the other to be revised. One needs some controversial stuff from moral theory and the theory of practical rationality to get the beliefs about justification and the motivational elements to interact appropriately in the first two cases, and this difficulty is absent in the third case. But even if one wants to accuse me of an error of rationality somewhere in my belief set, it's not at all clear that I'm unjustified in believing that p. Maybe the mistake is somewhere else, like in my acceptance of the normative principle.
I actually want there to be something wrong with believing that p when you believe that you lack epistemic justification for p or believe that you have epistemic justification for not-p, because there's an argument I'd like to build that relies on things going that way. But I don't think it's going to work, for the reasons in the third paragraph. So tell me why I'm wrong!