this problem took a while for me to see. The terms you discuss indeed don't automatically evaluate to zero. They actually make a giant, entirely positive contribution to the monopoles. I puzzled over this for ages until I had a conversation with james stewart. Terms such as the one below do not actually represent transition densities but static charge densities.
"SCF=XQC". Turns on the QC SCF procedure after the maximum number of regular SCF cycles have been done without convergence (if the SCF is converged in fewer cycles with the regular SCF procedure, then the QC step is skipped). The maximum number of regular SCF cycles by default is 64, and this can be controlled by "SCF=(MaxConventionalCyc=N)". The QC procedure is much slower than the regular SCF procedure but it is very robust and almost guaranteed to converge to an SCF solution. The "SCF=XQC" option thus provides a fall-back solution when the regular SCF procedure fails to converge. It also has higher chance to converge to the nearest SCF solution (which could be an unstable solution) than the regular SCF procedure, so it is strongly recommended to check for instabilities with a "Stable=Opt" calculation.