Dear Quan,Did you use TOCH or RHO for analysis? (after testing it on my own I see that TOCH is not suitable for analysis)
The main problem however usually comes from H (since PSEUDOpotential calculations cannot reproduce the charge on H correctly in principle, whether in SIESTA or any other program).
I guess, that changes in C in different environments (CH2 vs CH3 vs C-Ru) is a normal thing.
Alex.
