Image Center Position Coordinates in FITS Header Changed During Short Period Variable Run

Hello! I’ve been imaging the short-period variable, DL HER, with a period of about 12 hours.
For part of the several hour run, it appears that the position being input to the FITS header differs from the true position of the scope by up to 10 arc-minutes, and I am unsure how to find out why.
Questions: Could folks let me know how SGPro writes image center coordinates to the FITS header? Does the use of PHD2 and guiding affect how coordinates are written to the FITS header?
History: After about an hour during the imaging run, the image center location in the FITS header drifts down by about 10 arc-minutes from about 14:33 to 14:23. This drift may be gradual or abrupt. After about an hour, the FITS header changes back to the original location that matches the true image center.
When I try to plate solve the series in order to perform photometry, the images that are off by 10 arc-minutes do not plate solve. Visually, the images are nearly identical. When I compare an image that does not plate solve to the one that does, the centers differ by at most a couple of arc-minutes. I have to edit the FITS header to more closely match the true location of the image before the image plate solves and photometry can proceed.
I only run autocenter with plate solve when the scope first centers on the field. Thereafter, I run autofocus every hour.
Equipment and Programs: I use Pinpoint for platesolving and PHD2 for autoguiding. I have an LX200 8" classic, SBIG-402 imager with filter wheel and photometric filters, and Robofocus without temperature compensation. I use MPO Canopus and/or VPHOT (AAVSO program) for photometry. Canopus uses its own plate solver. VPHOT uses PinPoint. Both programs have problems plate solving when an image’s FITS header center is 10 arc-minutes away from the true center.
This drift in FITS header image center does not occur with the other short period variables I follow. I am unsure how to figure this out, and I would appreciate guidance. Best regards.

SGPro writes the RA and Dec to the header by fetching the position from the mount (as reported by the mount) when the image is saved. Please keep in mind that SGPro speaks J2000 only and both RA and DEC are recorded using it. This is not affected in any way by using PHD2.

Thank you! I am not sure why the object location is changed by 10 arc-minutes in the FITS header after an hour and then switched back about an hour later. I’ll try a run without PHD2 autoguiding and see what happens. Best regards.