So I did a little experiment, unfortunately the result just adds to the confusion.
The idea was that if I aim due south (180 degrees) at the horizon, the stars would appear to travel horizontally, and as a consequence the astrotracer would also move horizontally. If I aim to the right of due south, for instance 190 degrees the stars would appear to move slightly downwards. And thus slightly upwards if I aim to the left (170 degrees).
Since it is raining I just stayed home and placed the camera on a tripod (near a window to receive GPS signals) then aimed at a LED on the other side of the room. I made sure the camera was horizontal according to its own sensors. Since the LED is standing still I expected a straight horizontal line from the astrotracer when aiming due south.
Thanks to the dial with magnets I can fool the camera to believe it is aimed 180 degrees, and then start a trace. (20 sec with 600mm, it was actually 135mm but to get a little action I entered 600mm). The result was slightly downwards.
I adjusted the magnets to 172 degrees, which was to far and the line pointed slightly upwards. 175 degrees was the sweet spot with a perfectly horizontal line. So 175 degrees on the cameras compass results in a perfectly horizontal trace.
Now to the really confusing part, I adjusted to 355 degrees and expected a horizontal line again, but as in due north. But no, I had to adjust to 5 degrees to get a perfect horizontal line.
So when aiming south I had to remove 5 degrees, if I want 180 I enter 175. But when aiming north I had to add 5 degrees as if I want 0 I had to enter 5.
Confusing...