In other words, people will only convert for precisely the wrong reasons.

Absolutely true. If they were ready to accept the correct arguments, they would have become atheist on their own.

The probability for a given person to have developed a skeptical mind, have overcome the possible brainwashing effect of a religious education and all the possible correlated biases/fallacies (sunk-cost, belief-in-belief, etc.) and not having heard a compelling anti-religious argument is very low. Therefore, you can convert them either following the long, hard, but certainly more rewarding path of making them skeptical, if not rational, or you can bombard them with emotional nukes to demolish the emotional concrete walls that protect the religious belief.

People already have plenty of evidence; they just need a social reason to change their filter.

What I'd say is different between conversions and de-conversions is that the social incentive for religious conversion often takes the form of peer pressure or wanting to belong to a particular group, while de-conversion seems to just require evidence that you won't be completely cut off from good/cool/interesting people if you turn atheist.

I should say that the main evidence I have here is my own history (one data point in each set) as well as a few observed de-conversions in my friends (and one aborted de-conversion from a person who couldn't imagine letting go of his devout family and friends). And I've known people who converted (before I met them) to religion for reasons that sounded more like an innate yearning for deontology.

You can't talk people out of something they weren't talked into.