This still doesn't explain how you came up with the 2%. I've setup more carts than you on an order of magnitude and could not make that claim, nor could any individual.
Clearly, you've never one it. There's a shocker.
I'm just not pulling made-up factoids out of my hat.
It was amusing but definitely not funny to see objective measurements in German audio Press - of the then top Van den Hul cartridge.
With 50 dB+ separation in one direction, with less than 25 in the other. It was the costliest cart in the day.
Standard arm they used at the time was - you guessed it - SME V, which does not allow for azimuth adjustment.
If you have ever seen, let alone mounted the number of cartridges I did, to the degree of precision I require, you would have come to the very same conclusion(s).
I plan to write the detailed instructions how to adjust the cart using test records and an oscilloscope.
BTW, AT OC 9 and 33 PTG II are only specified to 30 dB channel separation. It has been long since I adjusted any of these (or their immediatev predecessors), but at least for OC Xs I do not remember they could exceed this 30 dB spec in real life - no matter what. AT does (did?) produce carts that could work better; I remember one AT-140ML that came to me with the complaint there is "something wrong with the balance". It was the record azimuth error, between 3 and 3 1/2 degrees, after correction that "crooked" cartridge registered solid 38+ dB, in both direction, phase of the residual signal being the same amplitude and phase both ways, across much of the spectrum, roughly to 10 kHz, where the usual tapering in high frequencies kicked in.
AT did produce one gem regarding ... - well, ANYthing - but I will limit myself here to separation. That cart not only does have very high separation in the midrange - but has, for all practical purposes, separation constant with frequency, at least to 20 kHz - the curve for separation largely mimics that of the frequency response, only reduced in level.
IF it is adjusted for azimuth.
Chime in once you are consistently capable of results as above.