Your SDE is better than mine in your long GA run but it is there and it is larger than your DEC seeing it seems.
Agreed, in the unguided ( and guided ) runs, the SDE is very clear, but PEMPro reports it as about 0.4 arcsec pk-pk now so not as critical, but certainly measurable ( and visible in the good seeing ).
That said, no "major" spikes or oversized oscillations are present in the guided data, so thats good. In fact it looks a bit like data from the very early 60ECs ie when unguided, there is drift and SDE
Guiding corrects the drift well, but only partially deals with the SDE.
( Comparing RA and DEC in the attached plot gives an idea of the difference between RA and DEC )
Again, the fastest frame rate here is around 2.1secs so may affect the data as its very close to the Nyquist freq for the SDE.
The unguided data would be nicer if you could collect at say 500ms or less
What's weird is your worm native PE seems to be showing through.
I disagree here ( at present ). I suspect what you are seeing is merely etching tolerances in the encoder grating
as if you overlap the data based on worm period, the data doesnt look the same for each pass.
ie as per above, the unguided tracking data of early models had SDE on a slowly changing baseline for the tracking rate, and the long slow errors changed based on sky position, hence suggesting it is mechanical tolerances in the encoder grid.
Have attached the unguided data run here plotted purely as raw with no drift removed
Shows very well the diffence in amplitude between RA and DEC in places,
Andrew Johansen Melbourne Australia