Somehow the third target CenA observations are ending up in the CenA-sky subirectory (FITS header + data does not match the directory)
This is likely to do with how POCS updates observations during take observation, since the field is based on the number of completed exposures in the first / primary camera
This is consistent with the problem being associated with the third exposure, as this is the last in the batch before changing to sky observations
Also consistent with the fact that the fourth observation (first sky observation) ended up in the right place
Possibly related to #423