I collected some data using a small refractor o YZ Boo (a delta scuti star). As you can see (below) the first part of the curve doesn't look right and I'm wondering why.
I've tried different reference stars and I even tried to put a simulated gaussian star and I always get a scattered first hour. Looking at the images individually doesn't show anything special (no clouds or obvious noise in the background).
I think I've narrowed it down to a CCD problem, some sort of noise that is non-uniform. Two nearby stars magnitude difference will vary by ~0.2 and are uncorrelated.
One more thing to note, I used 2x2 binning and the seeing was so good that stars sometimes fell within a single pixel. Have you seen this problem before? Is there a way to recover some of the bad data?