I try to use ATCOR and HCS resolution merge to preprocess urban Worldview-2 data. Applying the HCS algorithm without initial atmospheric/radiometric correction works just fine, but if I apply ATCOR beforehand, results get severely damaged. This happens:

Also, after using ATCOR the data also looks alright, and I can't seem to find the source for these distortions. If needed, I can deliver further information on the input data, histograms or whatever might help.