The following has the ring of truism: belief aims at truth, and preference/desire aims at the good. But Liz Harman argues against the latter claim on the grounds that loving a person as they actually are may lead us to prefer this actual state even to an alternative that would be better in every important respect. (Suppose you have a disabled child, now grown up. You love this person as they are, and so do not wish that their disability had been cured at birth -- even assuming this would have been better for them and everyone else too -- because then they would have grown up to be a radically different person from the one you actually know and love.)
My initial response is to simply bite the bullet and insist that love makes us irrational in these respects. It is not, strictly speaking, rationally warranted to be biased towards the actual, worse state. The better state is always preferable (it warrants preferring). It's just that the contrary bias is one we may be happy to have. It's fortunate that we prefer our loved ones as they actually are. Indeed, these biases are reflectively endorsable in a way that is familiar to sophisticated consequentialists: we do better to be biased in these ways rather than single-mindedly pursuing the good. But that doesn't mean the attitudes in question are rationally warranted (in the strictest sense). It just shows that sometimes we foreseeably do well to be irrational.
One conclusion you might draw from this is that we shouldn't be too concerned about having warranted attitudes. I think that's exactly right. My writing sample defended a form of 'rational holism', according to which we are instead advised to reason according to those norms which we can reflectively endorse. I thus agree with Julian Nida-Rumelin that it would be perfectly advisable to "refrain from point-wise optimization because you do not wish to live the life which would result." (He actually says it would be 'perfectly rational', which I earlier endorsed too, but for now I want to draw a sharper distinction between advisability and rational warrant.)
Further: if rationality and warrant and so forth are meant to be action-guiding -- concerned with reasons we can follow -- then perhaps we should simply identify these with 'advisability', and give up on the notion that preference aims at the good, as Liz suggests. That's compatible with a continued insistence that it's the impartial good which is the ultimate normative source. It's just that we end up with a neater and more coherent framework overall if we tie rationality more closely to reflective endorsement, rather than identifying it with the strict and alienating impartial ideal.
Anyway, I'm basically just thinking aloud here, and I'm not even too sure what my question is. (Perhaps: "What is the best way to think about rationality?") But if anyone else can shed some light on the issue, that would be immensely helpful...