I think it is not a bug… In the ewald or pppm code, the long range electrostatic interactions are calculated all at one time (total). It is not calculated for each atom seperately. Though it can be done but the code will become a little slow. I wrote my own script which calculates the E_long for each atom seperately after the simulations are done from the dump file (for ewald).
I could help you out if you want to know more. Let me know