Rajan,
Likely you have the habit, like me and I suspect most people, to
contour difference maps at +/- 3 sigma. In the beginning of
refinement, these difference map will have significant peaks at 3
sigma, i.e. positive peaks were there is nothing and there should be
atoms and negative peaks where the model has atoms but where there
should be nothing (plus perhaps the high noise peaks Eleonor, George
and Stephen mentioned). I.e. 3 sigma will correspond to quite a few
electrons per cubic Angstrom.
Once refinement is (nearly) complete, the difference map will be much
flatter, so the same contouring at 3 sigma will show up mainly noise;
or in other words, will corresponds to very few electrons per cubic
Angstroms.
Rather than worrying about the many noise peaks, do validation with
Molprobity or another program if you haven't already done so to see
how "good" your model is.
Or, if you really want to see an almost flat difference map, contour
using electrons per cubic Angstrom instead of sigmas. Coot mentions
both when changing contour levels.
Mark
