Below shows a short session where I used a simple Keypad Dec arc model to keep the star (Arcturus) on the crosshairs of my imaging camera. The keypad reads out the RA and Dec trim rate that is applied in the CP5 control in order to follow the star's motion over time. At the beginning of the run, at 7:52pm, the star was shown on the crosshair of the camera (right blue crosshair in MaximDL), while at the same time SkyX recorded the cursor position to be exactly aligned with the star. The trim rate being applied in RA was 61 arc sec per hour - this being the drift rate of the star at that 50 degree altitude, due mostly to atmospheric refraction and some slight polar alignment error.

After approximately an hour, the star is still on the blue crosshairs (mount's tracking was being adjusted continuously by the model). Pixel scale of this setup was 1.1 arc sec per pixel. You can see in the second image below that the actual reported position has moved by approximately 61 arc seconds in RA with respect to the planetarium position. If the mount had continued tracking this star for 4 or more hours into the West, the actual reported position would probably have been several hundred arc seconds. There is no field rotation involved, just simple and pure atmospheric refraction that causes a star to not be in the theoretical RA/Dec position that a planetarium program shows. Doing a Recal at any time will of course bring both the star and the cursor back together.