I agree with this idea wholeheartedly, unfortunately there are a few problems with documenting certain aspects of modding, specifically model making.
I for instance build models from a picture that I have in my head, there's no real 'concept art', maybe a few doodles to get certain areas of the model to look the way I want them to. I then proceed to use Rhino3D (not the most popular modelling program by all accounts) to build the model from first a three view wireframe (an X,Y and Z picture basically). I build the basic model shape face by face and then boolean the objects I want parts removed from or joined, I don't extrude/manipulate spheres and boxes like most of the other modellers. I also use multiple subobjects (which bob will give you a good telling off for).
This is a non conventional method of model building but it works and proves that if you do try to document a flexible process like this you'll have 6 different guides by 6 people who all do it in a different way depending on what modelling software they have access to and how they use it.
Documentation is excellent in theory but you just can't document some things, you have to get hands on
