The post and comments -- all very interesting so far -- tend to assume that a rationalist's goal in conversation with someone else is always a first-order one, where success and failure are represented only by whether the interlocutor changes his/her mind with respect to rationality or truth-seeking. There might be a limited category of cases where this assumption isn't good.

For example, if there were a young-earth creationist who was also a fourth-grade teacher, and who was in the habit of subtly undermining scientific truth with talk of "controversies" and the like, I think we could assume that he was erecting serious roadblocks to his students' future progress as rationalists. (I think I need not spell out the ways in which that might be true.) Now suppose I were able to persuade him that something like evolution by natural selection took place, even if I did so by some fundamentally dishonest means (by citing certain Bible verses, say), and in a way that led to no increase in his overall ability to think rationally about anything. If the result were that he kept his damn mouth shut about bogus scientific controversies in the presence of impressionable children, thereby not impeding their future progress as rationalists, the broad cause of rationality might well have been helped along.

Generalizing a bit, it seems rarely preferable to allow someone to continue believing something false that they have no justification for, instead of convincing them to believe something true that they have no justification for, in cases where -- as the OP seems to suggest -- those are the only two realistic options. To say we must never consider using some "dark arts" in the service of rationality at a level or two of remove seems somehow ... deontologist.