Well, I'll acknowledge that you could change premise 2 into an inference rule. But notice that you could change either premise 2—the pro-desire-satisfaction one and the pro-desire-frustration one—into an inference rule. Indeed, you could change any normative claim into an inference rule: you could change "people who want to have gay sex ought to go see a trained Baptist minister to get cured" into an inference rule, and then validly go from "I want to have gay sex" to "I ought to go see a trained Baptist minister to get cured". So from the fact that premise 2 could be changed into an inference rule, I don't think anything follows that might jeopardize its status as a full-blooded normative claim.

On the second point, I thought lukeprog was discussing direct conceptual reduction. But if he wants to provide hypothetical imperatives with a synthetic reduction, he'll need a theory of reference capable of explaining why the normative claim turns out to make reference to (and have its truth-conditions provided by) simple causal facts. And on this score, I think hypothetical imperatives and categorical moral imperatives are on an equal footing: since reductionist moral realists have a hard time with synthetic reductions, I would expect reductionist 'instrumental realists' to have a hard time as well.