Why am I getting substantially different energy values when

[gcates]

I am investigating the potential energy map of a periodic array of water molecules as a function of rotation. As part of that, I am attempting to determine the partial charges (to determine the dipole) using DDEC6 via Chargemol as a post-processing step (i.e. I already have converged WAVECAR files).

In order to implement this method, I need to include the core electrons and produce the relevant AECCAR* files by running VASP with the tags "NSW = 0; LAECHG=.TRUE." in my INCAR file according to numerous tutorials/discussions. In doing so, an energy relaxation is still performed, and I am getting different energy values (as show in the attached figure).

In the figure, the red line represents single-point energy calculations for different rotations. (Here, 1 rotation = 0.1 rad.) The blue line represents the single-point calculations when running VASP again with only the above mentioned INCAR tags as a post-processing step. Aside from their absolute value difference, qualitatively the potential maps differ as well.

Is this suppose to happen? Should I get such drastic energy differences when including core elections? On the one hand, I can see how explicitly including the core elections in a wavefunction (i.e. WAVECAR file) that previously omitted them would require some relaxation, but then this also suggests that the use of pseudopotentials is not justified (an obvious contradiction).