I was starting to look at the spectrum of the LED
Good point! MGIV has three emulsions since 1984; see:
and possibly the response of the "medium-hard" emulsion falls in a hole of the LED spectrum. See:, e.g.
Then you are left with only the two extreme emulsions, which makes more evident the inherent problem of variable-contrast paper, when the curves of individual emulsions don't merge well into the equivalent of a single-grade paper.
So much for my (vague) plans to try an LED lamp in my enlarger.