Cadaver, I meant that the software publishers have "almost got it worked out". We're not working on it at all. We do do some CAE in 3D, but for all it's fancy finite element generation and dynamic analysis, we find that the programs are cumbersome and demand compromises to make their use economical. (Or in some cases even possible.)
As for the "highly schematic" v. "dimensionally accurate" thing you have a point there. Each has it's own place in our work, and for some purposes we do "cheat" so that things will resolve at the chosen scale. On the other hand, plans are drawn with accurate dimensions where other dimensions need to be generated from those plans.
A very typical situation with clients is where they've drawn a wall to nominal thickness, and one side is the exterior face of the building, but on the other end of that building the opposite side becomes exterior. Obviously you cant give an accurate (and associative) dimension to both sides of such a wall if it's thickness is only nominal.
Most of our clients are of the opinion that the "mistake" is too little to worry about, as "they work that stuff out in the field." But it makes it dang hard to do geometrical constructions on a plan if you have that sort of inacuracy.
OTOH There are certain things we never bother to draw to scale. What's important is that the function be clearly displayed once the drawing is plotted. I don't care what size a nail really is, but if it spikes two pieces of lumber together it better look like it does on the plot. If I have to exagerate the thickness of a shim so it shows up, no problem, so long as the function is clear. The location of a ledger is cheated on a plan so it doesn't "blend" with the wall. This doesn't matter since everyone knows a ledger goes directly on the wall. (The location of the wall, however, must be accurate to avoid compounding errors as the drawing geometry is constructed.) In such respects, our drawings are "highly schematic".