I think that phd multistar has improved performance of guidescope guiding, but not OAG guiding. This is because fewer stars and a narrower field of view reduce the possibility of averaging errors. I've been getting similar results as you with a guide scope -- 0.4 to 0.9. When I did use the OAG, I found lost stars, and confused centroids caused issues. My usual tweaks were to monitor the star profile and tweak the gain, exposure and star choice until the centroid crosshair looked stable to my eyes. -- also make sure the OAG is in front of the main cameras filter, and use an ir filter on the guide camera.
Jamie Amendolagine