I got the impression that you aren't allowed any self-harm or evil acts. If you won't stop something for epsilon evil, then you care about it less than epsilon evil. If this is true for all epsilon, you only care an infinitesimal amount.

This sounds right to me, so long as 'self-harm' is taken pretty restrictively, and not so as to include things like costing me $20.

In his discussion of the 'murderer at the door' case Kant takes pains to distinguish between 'harm' and 'wrong'. So while we should never wrong anyone, there's nothing intrinsically wrong with harming people (he grants that you're harming, but not wronging, the victim by telling the truth to the murderer). So in this sense, I think you're right that Kantian deontology isn't worried about suffering in any direct sense. Kant will agree that suffering is generally morally significant, and that we all have an interest in minimizing it, but he'll say that it's not immediately a moral issue. (I think he's right about that). So this isn't to say that a Kantian shouldn't care about suffering, just that it's as subordinate to morality as is pleasure, wealth, etc.

I judge an ethical system based on what someone holding to it must do, not what they can.

It seems to me arbitrary to limit your investigation of ethics in this way. The space of permissibility is interesting, not least because there's a debate about whether or not that space is empty.

then the most important thing is that you, personally, are moral. Everyone else acting immoral is an infinitely distant second.

Agreed, though everything is an infinitely distant second, including your own happiness. But no one would say that you aren't therefore passionately attached to your own happiness, or that you're somehow irrational or evil for being so attached.