The plots below compare the whole CHICOS data set (red) to the showers just in the range 18.2 < LogE < 18.8 (blue), where there is a double peak in the energy distribution.
In these and other plots I did not see any characteristic that would allow us to filter out the double-peak. It is possible that there is a difference in the quality of the reconstructions in terms of number of sites, etc, but looking at that will require modifying the autoFit program to keep track of such things.