What do we want from a theory of desire? Should it match our intuitive judgments -- are we just analyzing the folk concept of 'desire'? Or are we aiming for a theory with some predictive/explanatory power? Or is it meant instead to have some kind of normative significance, so that fulfilled desires are pro tanto good for you? (These three options might lead us to focus on affect, drive, or evaluation, respectively.) Whether we opt for one of these, or something else entirely, seems to make a difference for how we deal with so-called "instrumental desires", for example.
Suppose I see a plastic apple, and - mistakenly believing it to be real - feel tempted to eat it. Should a theory of desire yield the result that I desire to eat it? Intuitively: sure. Normatively: no way. Behaviourally: whichever. (It's presumably just as explanatory to say that I desire some ultimate end -- a yummy taste, perhaps, or good nutrition -- and mistakenly believe that eating the plastic apple will serve these ends. This combination of attitudes suffices to explain why I might try to eat the plastic apple. If anything, it probably does a better job of explaining why I will stop eating it as soon as reality impinges itself on my beliefs!)
I'm most interested in the normative project, and this leads me to think that ultimate (non-instrumental) desires are the only desires we should count. Suppose I want to break out of prison, and believe that a hacksaw could help me achieve this end. Compare the following situations:
(1) I get the hacksaw, but it proves useless, so I remain imprisoned.
(2) I get the hacksaw and escape.
(3) I simply escape (no hacksaw required).
Surely the right thing to say here is that I get all that I want in situations (2) and (3), whereas in situation (1) I don't get what I really wanted at all. It's not as though I can console myself with the thought, "Well, at least I got this hacksaw I wanted!" I don't really desire the hacksaw at all; I just wanted to get out of prison. Similar problems arise when comparing (2) and (3). It's not as though I get more of what I want, in any interesting sense, in case (2) -- that would be double-counting! As an escapee in case (3), I won't feel the slightest frustration at having my 'desire' for a hacksaw remain unsatisfied.
So it is only ultimate desires that are philosophically interesting, I think. Sure, one could use the term in such a way that one counts as having a 'desire' for the believed means as well as the end. But what's the interest in that?
Another example: Liz Harman suggests that many apparent conditional desires, e.g. to be a fireman when you grow up, are really unconditional desires that simply weren't thought through all that carefully (i.e. one took the rationale / apparent 'condition' -- that you still want to be a fireman as an adult -- for granted). Maybe this fits better with folk intuitions, or a natural way of talking, or some such. But I don't really see why that should matter to us. Treating the desire as conditional gives us a neat explanation why it doesn't impact one's welfare when the condition isn't met. What does the alternative account give us?