Sanity and grace ...
After a somewhat sleepless night, and a distracted day, I went over the magnification exposure correction calculations and measurements and found a pair of dumb errors that caused the results to get skewed.
The (M+1)^2/(m+1)^2 formula and the metered values all agree within a few hundredths of a stop -- which is what they are supposed to do.
The 'square law', beloved of all 7th grade science students, does still work. The law doesn't work on magnification but on distance from the light source to the screen. In this case the light source is the illuminated aperture of the lens and the screen is a point on the photographic paper (or you can turn it around, projecting a point on the paper onto the area of the aperture). The DA magnification exposure correction ruler works on this principle. An error in it's use can arise if the position of the lens' apparent aperture isn't located properly.
The geometric forms - the ruler and (M+1)... formula - are based on thin lens formulations. The actual light fall off will deviate slightly as the size of the front entrance pupil - the effective aperture - changes a bit with lens to paper distance. You can see this by looking into a wide-open lens at several distances. In practical terms this slight error can be ignored.
The meter has one great advantage - no math and transcription errors.
The horse is finally dead, the flies have gone.