Some problems stem from representating a human state of mind as a static object, whereas our beliefs aren’t constant w/r/t time spent in reflection.
Every desire can be chained iteratively until you reach “the good”, but almost never is that entire chain ‘in register’ for real human being. At instant t in time, we are cognizant of a finite number of steps along that chain (we implicitly accept some goals as final, even with more thought power we would view them as instrumental).
Therefore the eliminitivist is not susceptible to Collard’s criticism. Whether the man’s desire was satisfied would be dependent on his mental state while having the desire. If he did not invest enough effort into chaining the desire all the way to The Good, then it’s possible for the desire to be satisfied but the person not (and more likely the fewer links along the chain are ‘in register’).
Collard’s “object”/“aim” structure of a desire seems to be just considering two points along the chain - this is still more nuance than a one-point approximation, but it really is referring to a pair “naive desires” as a desire with bipartite structure.
It is a reasonable strategy to salvage the “separatist” POV by giving an account for the relation between different desired states of the world (they are not atomic anymore)