(1.) Philosophical ethics improves (or selects for) moral reasoning.
(2.) Improved (or professional habits of) moral reasoning tends to lead either to (a.) better moral knowledge, or (at least) (b.) more frequent moral reflection.
(3.) (a) and (b) tend to cause better moral behavior.
Therefore, ethicists will [tend to] behave better than non-ethicists.
But ethicists don't tend to behave better than non-ethicists. So should we jettison (1), (2), or (3)? I'm inclined to think that the question is misleading; what we should jettison is the presupposition required to get from (1) + (2) + (3) to the conclusion: that ethicists do philosophical ethics.
That, of course, is the paradoxical way to put it. What would be slightly more accurate would be to say that most ethicists don't do very much philosophical ethics of the kind that examines 'moral reasoning' of the sort that leads us to expect moral knowledge and/or moral reflection relevant to behavior. As I've pointed out before, when you look at what is usually taught in ethics classes, it doesn't have much to do with practical issues of moral reasoning; what is taught is rather abstract -- speculative reasoning about how best to reason about the normativity involved in moral reasoning (and that moral reasoning itself is often, although fortunately for everyone's sanity, not always, removed from the sort of moral reasoning you'd genuinely find in actual moral life). We can consider it ethics in some sense, and it has its place in moral life, since it gives people some general tools useful for categorizing types of moral reasoning that they may find, either in private life or in the public sphere. But while it has a definite use, it's surely a very limited contribution to actual moral reasoning in practical life. And in ethical research the level of abstraction, and the minuteness of its relevance to actual moral reasoning, is sometimes even worse. We wouldn't expect physicians to be better physicians merely because they know possible types of diagnosis; that has its use, but merely as an instrument for learning and communicating matters of actual medical work and practice. What improves medical reasoning is the teaching of actual medical reasoning, not merely the teaching of some of the elements that make medical reasoning easier to categorize.
So one can really say that 'moral reasoning' is equivocal in the argument; there is a sense in which each of (1), (2), and (3) are true. But it is not the sense in which contemporary ethicists have much to do with moral reasoning.