Hello Royaloak,
The reason for the suggested undercompensation is the usually higher than acceptable sidethrust on the cantilever, caused by an antiskating mechanism. While the force excerted by the stylus/diamond on both groove walls may be equalized, it causes a permanent cantilever displacement( since the antiskating acts on the arm, not the diamond itself), resulting in a non linear behaviour.
Imagine a speaker being exposed to a certain amount of a DC voltage. The null position of the speaker(where it rests when no signal is present) will have moved forward or backward, depending upon the polarity of said DC voltage. The maximum cone displacement is reached sooner in one direction, it takes a different amount of force to move it forward vs. moving it backward.
Same for a cantilever that has a conventional suspension, dynamic gradation is foreshortened and the resolution of microdetail(what happens around the Null point) diminished.
If one is to reduce the skating compensation to zero, the non linearities caused by non-equal pressure of the diamond on the groove walls will have a larger influence, so zero antiskating isn't the answer either(imho).
And I have not even touched the issue of the coil position against the magnetic circuit when a preload is present. Not as severe an issue and highly generator design dependent, but...
As said in the manual of my arms, my suggestion is a starting point, not an absolute(ly perfect) setting. If your method yields better results, - or results you like better, then, by all means, go by that!
All the best,
Frank