Right. Sorry, I don't know why your reverse RA is not working.
...except that I just noticed you seem like you're set up for ST4 guiding. Is that intentional? You should really have your mount listed as the guider
Do things improve if you change the "Guider" in your optical train from the QHY CCD to the iOptron mount?
For the future, if/when this problem gets solved I would suggest more than 3 max iterations (e.g. 10) and more max-move pixels (e.g. 25).
Your PHD2 calibration plot shows that it used 12 iterations and >25 pixels out.
If the above doesn't fix things and you want to debug more, you might try other parts of the sky to see if it is somehow related to telescope weighting. Also, you might see what "Calibration step-size" PHD2 is using, which is probably the equivalent to the "Pulse" input to the Ekos guider, and try that. However, 1000ms seems reasonably long to me, so I'd be surprise if you needed more. On the main Guider page, please make sure that all the direction boxes are checked, and that the telescope specs seem reasonable