I agree that it is a more difficult task to create a quality IEM that has multiple drivers versus one dynamic driver. For that reason, I would say that in the majority of cases, given a particular price-point, the dynamic driver will win out simply because the engineering challenges, as you suggest, as fewer. However, you do reach a certain point in which, for lack of a better term, one runs out of headroom. Making a single dynamic driver that competes at the $1000 level, for example, is much more difficult than making a multi-BA that does so. There are simply limitations to what any driver arrangement can do, and there are numerous areas in which one driver type has less of a limitation than another driver type.
Additionally, while issues with phase or crossover are problematic, they are becoming less so every day as more and more companies deal with the tech.
If you compare the worst of one type to the best of another per your analogy, it will yield much different conclusions than if you compared average to average. An average car is going to be better in the vast majority of use-cases than an average horse and carriage. I think it's possible to make similar claims regarding driver types.