Shortly mentioned here , but maybe I rather put it in its own thread.
The center coordinates that the Mosaic tool of the scheduler creates are wrong (unless the mosaic is at the equator): The RA computation just assumes one RA second is 15 sky arc sec, but that is missing the factor 1/cos(DEC) and the overlap ends up being much to high.
E.g., for a 128x97 arcmin field with 5% overlap I get center coordinates in DEC 60⁰52'10" and 60⁰20'24", this is a difference of 92', which is 95% of 97' -> correct.
But for RA I get coordinates 2h58'36" and 2h50'31", ~8' in RA. At the equator this would indeed be 120', 95% of the width, but at DEC 60⁰ this is only 60'

I just pushed a possible fix. pixelsPerArcmin was splitted to pixelsPerArcminRA and pixelsPerArcminDE with the former accounting for the cos(dec). Unfortunately, didn't really have time to test this.

It was a simple fix (or so it appears). Please recompile from GIT and check. It might still break at higher declinations.