I think this might be making the same kind of error that I expect the Equal Vote Coalition is making: ignoring that things become more complex in higher dimensions, and simple symmetries may not appear where you expect them. In this case, as more candidates emerge, we see the possibility of Condorcet cycles,and your seemingly obvious criterion becomes impossible to satisfy.

Don't get me wrong: I think Condorcet efficiency is an important goal to pursue. But your stronger condition, that we should never elect a candidate who loses in any head-to-head matchup, is just impossible to satisfy in general.