[molpro-user] Question on MRCI energies in multi-state calculations

[Lorenzo Lodi]

I noticed that when doing MRCI calculations (CASSCF reference) the 
energy of a certain state is different if that state is computed by 
itself or together with other, higher-lying states.
For example, in this mock calculation on CO:
geometry={C; O, C, 2.2}
{RHF}
{CASSCF}
{MRCI; STATE,1}

I get
RHF=-112.743047
CASSCF=-112.879927
MRCI, STATE,1 = -113.047810
MRCI, STATE,2 = -113.047936
MRCI, STATE,3 = -113.048093

The multi-state energies are consistently lower than the single-state one.
I also get a decrease in energy only for the multi-state MRCI 
calculations when doing the calculation without point-group symmetry 
(NOSYM keyword):
MRCI, STATE,2 (NOSYM)= -113.047896
MRCI, STATE,3 (NOSYM)= -113.048010

Of course these energy difference are rather small (10-60 cm-1) but I 
would like to understand for what reason they exist at all.
Any hints?

Lorenzo