Hi guys,
the analysis of H2O is meant to be wrong with SIESTA since it uses
pseudopotentials and it will never produce the peak of charge on H.
In most of other cases it will work fine even with RHO (without any core
corrections).
If you really need to deal with H (or other atoms without charge peak on
atomic center) you can add an artificial Gaussian of 1e charge on it
(for H, or more for other atoms) with a very small radius not to affect
the rest of charge distribution (don't forget to subtract that 1e after
analysis).
The grid required by the UTexas code is ~300Ry.
Alex.
