I'm pretty sure this has been discussed at length in the RST-135 thread
Part of it has.
The thought of finding a sweet spot and rotating the spline drive to match it is new and interesting.
Anyway, there was even a link to an academic paper analyzing, in the context of strain wave reducers, the PE dependency to actual load and angle.
Oooh do please find the paper
I have seen several papers on the topic, but in all cases, the test rigs had rigid axles on each side
then slotted in a harmonic drive.
This removes the longitudinal bending moment effects that come about when driving
a cantilevered load.
As i mentioned in the other thread, i can visualise the different loadings in the flex spline and cross bearing
but how it results in an error in PE and swash ( at the arcsec level ) is beyond me.
Andrew Johansen Melbourne Australia