The new version of Dmol3 provided with Materials Studio (v3.2) has a “new” bug that did not exist in the Cerius2 v4.9: The ESP charges calculated with a pseudopotential gives awesome results, like charges of +5 or -5 per atom ! The calculations are fine when NOT using any pseudopotential. This bug seems to have been fixed in version 4.0
Vamp is a semi-empirical quantum mecanics program provided with Materials Studio (v3.2). The ESP charges seems to depend on the origin of the coordiates, ie the first atom. This strange feature is probably related with the polar representation of the atoms. This bug does not exist in MOPAC7 (sourceforge version 1.1). In facts I was not even able to reproduce the charges published by the authors Beck, Clark & Glen; Journal of Computational Chemistry Volume 18, Issue 6 , Pages 744 - 756.
This (free) software is painfull to install (because it needs the latest libraries, especially for OpenGL), nevertheless it seems really promising !
I compared a hi-resolution XRPD taken at ESRF ID31 with the powder pattern simulated (from single crystal structure) by some programs. The best result was produces by Platon (Spek et al) far better than Cerius2 and Materials-studio from Accelrys. The worse results were obtained by Mercury 1.4 and 1.4.1. Mercury doe probably not use disordered populations.