I think Markus refers to the fact that if you remove the generated table, and sync again, still just 1 field is added for the subtypes and all subtypes map to that field, instead of getting their own field.
The designer follows a simple rule for this, which is the same as when you map multiple entities onto the same target: it simply processes the entities as if they're mapped alone on that target. The reason is mainly that the alternative would lead to micromanagement settings where you have to specify which entity leads the table shape: is it the supertype, entity A, or maybe A and B are equal and C's field requires a different field...
There's no real solution for this, so we left this as-is.