However, my thoughts are that you're courting the nature of the fall-off at the edge of the excessively large light mod (not just the rate of [dictated by distance], but the quality of [dictated by relative size, of which the larger mod is always going to win if working distances remain the same])?
Not really... because the "apparent/visible size" of the modifier is not really "larger" when feathered like that. Take a plate and look at it from a 45* angle... it becomes oblong in size. As far as the area being lit is concerned, the apparent/visible size *is* the size and therefore has the associated "quality." There will be a different effect elsewhere due to changes in distance, but they will primarily affect areas of no concern (i.e. below the waist).
Assuming even fill of the modifier (and that *is* a large assumption in many cases) there should be no difference in light "quality" across the surface.
At the very edge there *is* less light and less wrap, but that has to due with being at the edge and not size per-se (but it tends to more gradual than a flag is). Both of those effects would generally be considered "lower quality" (less soft).
"Hot spots" are due to either uneven fill, or falloff due to having it very close. Uneven fill is rather common, even with studio heads (but to a much lesser degree). In fact, feathering a modifier like that is often done *because of* uneven fill in order to take advantage of hotspots.
IME/O, most people do not use modifiers from close enough in general, not to mention close enough to cause really dramatic falloff (hotspots).
All modifiers of the same design/shape will light the same size area w/ very similar "edge falloff" characteristics when used from the same distance. "Size" primarily affects "wrap" (fill) which occurs towards the center of the illuminated area.
Edit: IMO many do things simply due to the results it derives, which is fine because nothing else really matters. But they frequently do not really understand the cause/effect relationships. And that frequently leads to miscorrelations and "bad information."