(A) If God knows counterfactuals of freedom, He knows it in virtue of knowing His nature or His choices.
Generally Agreed Premise 1: He does not know them in virtue of His nature.
Generally Agreed Premise 2: He does not know them in virtue of His choices.
Therefore, God does not know counterfactuals of freedom.
The Molinist denies (A). The anti-Molinist asks how else He would know it. The only answer Molinists have ever given is the supercomprehension answer: God just really, really, really knows free creatures. But this is just handwaving, and does not appear to yield the necessary result. Molinists have therefore given no reason to think there is some third way God knows things.
Further, we have reason to doubt that in the relevant cases there is anything to know.
(B) Something is knowably true only if (a) it happens; or (b) its causes are such that it, and it alone, will happen.
Generally Agreed Premise 3: Nothing corresponding to (a) is found in counterfactuals of freedom, because they are counterfactual.
Premise 4: Nothing corresponding to (b) is found in counterfactuals of freedom, because they are of freedom.
Plantinga, at least, seems to have committed himself to denying Premise 4. He gives no reasons for doing so. Most Molinists, I suspect, would instead deny (B). We are still waiting for a reason to think anything else is knowably true.
The problem, you see, is that Molinists attempt to evade the grounding objection altogether; but one cannot both evade the grounding objection and give any direct reason for thinking Molinism true. A direct reason for thinking Molinism true (i.e., a reason that isn't based on seeing how useful it is if it is assumed to be true) would have to answer the grounding objection. Because Molinists don't do that, they give no direct reasons for the truth of the middle knowledge thesis; the only reasons they give are reasons for regarding it as a useful hypothesis (it solves problems if it is true). But, unless the grounding objection is answered, there are direct reasons for taking a non-Molinist position (e.g., the ones given above). And therefore, without an answer from Molinists to the grounding objection, we have reason to think the non-Molinist position is not merely a useful hypothesis but is actually true, and no reason to think that the Molinist position is anything more than a useful hypothesis. That is, the strongest we can say in favor of the Molinist position is that we have reasons to think that if it is true, it solves problems; the strongest we can say in favor of the anti-Molinist position is that we have reasons to think that it is true.