From the examples the people I mentioned above came up with, guilt seems to be a very strong trigger of rationalisation.
Perhaps rationalization is an adaptation that develops when people risk some kind of punishment for their irrationality.
We are irrational, and we already suffer the consequences of our irrationality. But if there is additional penalty for admitting irrationality, it gives us incentive to pretend that the irrational decision was in fact rational; to lie to others, and ultimately to lie to ourselves. Admitting irrationality can be a very bad signalling.
How exactly does guilt become a part of the equation? Probably by believing that there is no such thing as irrationality, and people are always perfectly following their utility function. So if you forgot to do something, it means you decided not to do it, because it gives you negative utility. So whenever your irrationality harms people around you, it means you hate them. (If your irrationality sometimes harms you, this can be explained away by saying that you didn't really care about something, only pretended it.) From the outside view, our irrationality is not credible -- it may be just a public act, while we are following our true preferences (defined circularly as "that what we are following", plus some possible secrets).
Anna Salamon and I are confused. Both of us notice ourselves rationalizing on pretty much a daily basis and have to apply techniques like the Litany of Tarski pretty regularly. But in several of our test sessions for teaching rationality, a handful of people report never rationalizing and seem to have little clue what Tarski is for. They don't relate to any examples we give, whether fictitious or actual personal examples from our lives. Some of these people show signs of being rather high-level rationalists overall, although some don't.