[In standard cases of 'rational irrationality'] the action is irrational because of its harmfulness but there are reasons for being in a state that leads to the action, since being in that state has benefits. The cases we now consider are different. The reward is not tied to our wanting to have a desire for the saucer of mud or to our trying to have it. It is dependent on the desire itself. Similarly, in the paradox of hedonism, the reward is tied to favoring other things apart from pleasure, for their own sake, and not to wanting to have these pro-attitudes or to trying to have them.
It's not obvious that this is a relevant difference. In their case what's desirable is still not the object of the desire -- the saucer of mud itself -- but merely the state of desiring it. So it still seems natural and accurate to describe this as a case where it may be rational to bring it about that you possess an intrinsically irrational attitude (in this case, a perverse desire for mud). But even if one disagrees with this classification (and I can't imagine why one would), bringing up those standard cases of 'rational irrationality' at least serves to remind us that this is a perfectly coherent way to describe the situation.
R&R then offer a common (and, I'll argue, misguided) argument from analogy in favour of state-based reasons for desire:
Surely, if we are supposed to have reasons for actions when the actions have useful effects or are valuable for their own sake, why shouldn’t we have reasons for attitudes in comparable circumstances? The two cases appear to be perfectly analogous. Pace Gibbard and Parfit, it is therefore plausible to conclude that we do have reasons for beneficial attitudes, even when they are directed to objects without value. To be sure, we also have reasons to want to have such attitudes and to try to have them, but this is because we have reasons to have them, in the first place.
I think there's a natural answer we can offer to their rhetorical question. First note that, in case of attitudes, we can distinguish between the content (or object) of the attitude, and the attitude itself (i.e. the state of possessing the attitude). This explains how gaps between fittingness (object-based reasons) and fortunateness (a state-based evaluation) can arise for attitudes like belief and desire. But this distinction collapses in case of actions: there's no gap between the thing to be done, and the doing of it. Since these coincide in one case but not the other, we need to be careful when drawing analogies between actions and pro-attitudes like desire. Compared to a fortunate action, is an analogous pro-attitude one where the object is likewise fortunate, or one where the state is likewise fortunate? Each is analogous in a different way, so it seems overly hasty of R&R to just assume the latter answer.
Here's the situation. We know that we have reasons for action when (i) the thing to be done is a beneficial thing, or, equivalently, when (ii) the act itself -- the doing of it -- would be beneficial. Now we wonder when we have reasons for desire. It's uncontroversial that we have object-given reasons for desire, i.e. reason to desire things that would be beneficial. That is, it's uncontroversial that type-(i) considerations give rise to reasons. And since (i) and (ii) are necessarily coextensive in case of actions, that suffices to explain why it is that we have reasons for action whenever the act would be beneficial. We don't need to further suppose that type-ii ["state-given"] considerations themselves give rise to reasons. That would be a completely gratuitous assumption. So there's no basis here for concluding that there are state-given reasons for desire.