If the program complains that you have negative intensities, this is usually due to misidentifying star observations as sky. It can also be caused by marking sky observations as star data. Check the STARSKY column of your data tables.
Another problem with sky occurs when the program cannot find a suitable sky observation to subtract from a star datum. This can occur when different measuring diaphragms have been used. Remember that you cannot correct an observation for sky unless there are sky data taken in the same band through the same aperture.
Finally, as the sky-modelling algorithm does not try to model twilight, it uses the ``nearest-neighbor'' method to correct star observations made during twilight for sky. You may find that such observations cannot be corrected for sky by this method, if the star was measured during twilight, but the corresponding sky measurement occurs just after the end of evening twilight or just before the start of morning twilight. You can prevent this problem by measuring sky both before and after each star measured during twilight; but you should do this anyway. Note that the planning program tells you when twilight begins and ends.