In 'The Normative Irrelevance of the Actual', I explained why it doesn't matter whether a putative counterexample to a moral theory is actual or hypothetical in nature, on the grounds that first-order moral theories can be understood as (implying) a whole raft of conditionals from possible non-moral circumstances to moral verdicts. But there's another, perhaps more intuitive, way to make the case, based on the idea that some counterfactually superior moral theory should be superior, simpliciter.
Consider Slote's sentimentalism. According to Slote (2007, 31), wrong acts are those that "reflect or exhibit or express an absence (or lack) of fully developed empathic concern for (or caring about) others." The relevant kind of empathic concern is not some kind of a priori theoretical posit, such as universal love, but rather is tied to our actual natural dispositions to favour those near and dear. (This is crucial to secure his desired anti-utilitarian verdicts.) But this raises the obvious worry: what if our "natural" empathic dispositions turn out to have racist or otherwise clearly immoral built-in tendencies?
Slote responds: “The ethics of empathy may here be hostage to future biological and psychological research, but I don’t think that takes away from its promise as a way of understanding and justifying (a certain view of) morality.” (p.36)
But, I suggest, if we know that there is a possible situation in which sentimentalism is not the correct moral theory, then we can ask ourselves what the correct moral theory in that situation would be. And once equipped with that correct possible moral theory—one that provides an independent justification for rejecting racist or otherwise immoral sentiments even when sentimentalism cannot—then we may wonder what we need sentimentalism for. What is stopping that counterfactually correct moral theory from also being the actually correct moral theory?
Perhaps there are some moral theories that give plausible verdicts only in a certain counterfactual world, and are no longer plausible when we apply them to our world (or others). So, fine, discard those clearly inadequate theories. Still, given the entirety of logical space to choose from, we should be able to find a theory that yields the desired results in our world as well as in the counterfactual world where it is superior to sentimentalism (or whatever merely contingently plausible theory we are considering).
So, if a moral theory is merely contingently plausible, we can find a better option out there. Being possibly wrong, in this sense, suffices to establish that the moral theory is actually wrong.