"""""""""The fact is most cam systems even now compute a toolpath based on a tesselated model derived from a nurbs model. Therefore, even though I may give a vendor a nurbs model, their machine is cutting it as G1 or straight line moves. So when they complain that they can't machine from from a triangulated model, that points out me they actually don't understand their own machining process very well.""""""""""

Well, yes to what you said, but:

Forcing the CAM program to generate it's path based off of the given little flat lines and patches and hundreds of thousands of edges (each requiring the calc to start and stop THERE as opposed to where IT DECIDES) is problematic for CAM programs and post processing into a proper toolpath. Some of your vendors may have just decided that the cost of problems that can arise because of that is not something they want to keep dealing with.

It's far better for the CAM program to be allowed to generate it's own little sliced up model based on the tolerances supplied within the system. Mixing those tolerances with the "forced dice" of a polygon model is hard at best.

Many CAM programs handle toolpath differently. Some will calc off of the "display mesh" created (way junk). These are the apps that will put a toolpath on a multi million poly model "really fast"... Some will just give you a set, hard coded tolerance which can burn past any of the really hard calcs. Most of the higher end stuff, will be looking at the surfaces supplied.....

CAM'ing a multi million poly model out when the specs are a little looser like you mention, is not super difficult, but it requires an understanding of the nature of the model. Not a lot of CAM shops are very versed in the underlying structure of the models they are working with. They are Machinists....