Earlier quoted context omitted.
However, that's not how the paper defines variability. I did think for a second that your definition was the case, and variability was a function of phenotypical expression, but it (as defined in the paper) is strictly about the statistical distribution of desirability within a population. See p4.
That's how variability is defined, yes, on groups (or, technically, on "probability measures")—which might be the entire population or a subgroup thereof. The term "subpopulation" isn't defined, and it isn't explicitly stated whether the high variability is a heritable trait. I believe your contention is that that the overall population has basically the same genes, everyone rolls the dice, and then the subpopulation…
"The actual magnitudes of these desirability values are assumed to have no significance, and are used only to make comparisons between individuals. Here and throughout, it will also be assumed that the same desirability value is assigned to each individual by every member of the opposite sex. "