I guess the thing for us is that, after utilising 3D on one large job, we have given up on the 2D+3D method, because I found that the parametric objects were too limiting. It's not so much that they don't do everything possible under the sun but that they either have small bugs or small limitations that are deal-breakers. Once you hit a bug or limitation that requires you to destroy the parametric nature of the object there becomes no point in using it, apart from as a design tool.
We'll use the stair tool, for instance, as a check to make sure we're calculating our stairs correctly but most of the stairs we do in multi-storey buildings simply can't be created with the stair PIO.
The window PIO is another example, with the sill bugs, and not only in 3D. When I last used it, it couldn't even represent a sill in 2D correctly either.
I think a lot of these problems stem from these PIOs starting their lives designed for the McMansion market. If NNA came at it from a different perspective (i.e. a global market of diverse architecture) I don't think we'd have all these limitations.