I considered that one, but there are two problems with it. For one, because so many of the garments of this general type are relatively loose across the torso/chest/belly, it would be hard to identify a benchmark for just how much fullness is the minimum necessity (since it's always going to be more than the anatomical measurement). But also, it would be very hard to identify a standardized method of taking the measurement. As a general rule, the narrowest point is going to be below the attachment of the sleeve+sleeve-gusset and above the start of any skirt gussets ... but depending on the construction this point can either fall across a broad band from armpit to point of the hips, or it may not exist at all (if the sleeve attachment is very deep, as it is for some albs).

I was worried enough about the shoulder measurement being more noise than signal, except that it helped provide a reality-check on where the overlong spans were coming from. I think I may go back and add in a metric for sleeve-length vs. height, since the shoulder data suggests that the body looseness is the primary driving factor for long spans. If I were dealing with more garments with strongly shaped armscyes then I'd have to worry about the difference between dropped and normal shoulders, but for the set of garments I stuck to here, there's relatively little shaping.

One thing I didn't mention in the above analysis is that at one time I starting making up an exact-size reproduction of the garment based on my article ... and rapidly noticed the same sizing issue in a very practical way. Hmm, it occurs to me that another non-invasive evaluation of the interpretation would be to take an exact-size reproduction to Paris and hold it up in front of the display case ....