My question was what level of theory the properties obtained from a (X)MCQDPT2 calculation correspond to. According to Alex, the properties are calculated at "zero order of QDPT", which is not the same as the CASSCF level. More precisely, CASCI vectors obtained from PT are used. These are however of zero order, that is they span only the CAS-space and do not include the external configurations. The vectors however account for interaction/mixing of the states. If the latter is weak then the vectors are close to the corresponding CASSCF vectors.