Higher-order coupled cluster calculations

Dear Mihaly,
dear staff,

checking some higher order coupled cluster calculations, which I
had done some time ago with older versions of MRCC, I noted
seemingly strange behavior:

The CCSDT(Q) is larger in absolute terms than the CCSDT[Q] energy,
and both differ from CCSDT. The same is seen for CCSDTQP(6) and
CCSDTQP[6] relative to CCSDTQP and this is what I do expect.

CCSDTQ(P) and CCSDTQ[P] energies, on the other hand, are the
same to machine precision, and both somewhat larger in absolute
terms than CCSDTQ.

I observed the described behavior for a range of very small
molecules.

Now I reran calculations for CH4 (cc-pVDZ) using a recent
version of MRCC (2017, statically linked) and reproduced this
behavior. See attached file.

Is there any reason to expect E(CCSDTQ(P)) = E(CCSDTQ[P])? After
all CH4 has 8 valence electrons to correlate and so we still have
some way to go to FCI.

Thanks a lot,
best regards,

Dirk