In the absence of differential flexure (and it would need to be variable), there would seem to be no reason why you should get variable results just from using the guide scope during PA. A couple more questions for you:
a. Can you describe which flavor of PA you are using (direct polaris solve, or rotations at off-polaris sky positions)?
b. Have you tried to do a drift align with Phd2 for comparison?

I had a similar problem with the Ekos PA software on my Celestron CGX-L mount (rotations away from polaris approach). I also experienced variable results, and lost faith / gave up on that software. A PHD2 drift align solved this variability issue for me, so I suspect the implementation in Ekos (vs the mount or other possibilities). You might consider at least trying a PHD2 drift align, followed by running the Ekos PA routine (multiple times) to do the same variability analysis as you've already done. That will be a good data point for you to consider.... CS Doug

Read More...