Might a Kantian improve the argument:
1) The forms of human experience are space and time
2) If abstracta / noumena exist, they exist entirely outside of space and time
3) ( (1) & (2) ) → (4)
4) Humans cannot have knowledge of abstracta / noumena because they are situated outside of human experience
5) If Platonism / noumenal realism is true, then ~(4)
6) Platonism / noumenal realism is not true.
It seems that we might conclude that transcendental idealism is the best we can do. This leaves open a heavy instrumentalism for AE, or a pragmatic justifications of AE.
Now, an obvious Sellarsean move is to deny (4) because it equates knowledge to some sort of empirical conception of knowledge. Just because noumena are located outside of human experience doesn't mean that we cannot know about them. An ABE or hypothetico-deductive argumentation might fit the bill. We then would still have room for a transcendental realism about noumenal entities, given that our scientific realism is radical and critical enough to allow it.
But, this being said, if we move to a transcendental realism which is sufficiently radical and critical, we may well face something like the original Benacerraf Dilemma, since such a critical and radical scientific realism presumably wants to still be in some sense physicalist about events and things. I certainly would want to resist Platonism, even if I deny the empiricism Kant uses to motivate (4). Platonic realism just seems to be plain old idealism to me.