Are you saying the OAG does not have to be in line with the CCD chip to work?
I'm not sure that I understand the question.
In order to work, the OAG's pick-off prism needs to be somewhere in the optical path where there is sufficient illumination and the stars are not too badly aberrated. Also, the guide camera sensor needs to be parfocal with the main camera sensor. This latter thing often results in some fiddling with the spacing between the filter wheel and the OAG.
Rotationally, the OAG doesn't care about the main imaging sensor at all. It can be at any which angle, and as long as you calibrate your guide software in that orientation, it will work fine. The main sensor might care about the OAG, though. Specifically, if the pick-off prism gets in between the main sensor and the scope, then the pick-off prism's shadow will be evident in the main image.
Often times, people will orient the OAG so that the pick-off prism sticks into the light path at the long edge of the main sensor. This lets you get the pick-off prism into the light path as far as possible without obstructing the main sensor. If you have a telescope with an imaging circle that is much wider than the main camera sensor, then you can have the OAG's pick-off further off axis. In that case, you can rotate the OAG independent of the main camera. This can be advantageous for scopes with a slow focal ratio, where you might need to position the OAG so that a bright enough star lands on the guide sensor.
I hope that this makes sense,