DETCI finds an "extra" excited state

Hello,

I would like to compute several excited states of imidazole at the CASCI level using the DETCI module. However, I found the energy of the fifth singlet A’ root is different from that obtained from MOLPRO, PYSCF, and FORTE. Here are the lowest ten singlet roots from PSI4 and PYSCF:

                 PSI4              FORTE              PYSCF
0   -224.691712103656  -224.691712104509  -224.691712094605
1   -224.429346130922  -224.429346135865  -224.429346112673
2   -224.390513013585  -224.390513011981  -224.390512997860
3   -224.320978021557  -224.320978022408  -224.320977987456
4   -224.295459370323  -224.292448261076  -224.292448213622
5   -224.292448260642  -224.246878466006  -224.246878429002
6   -224.246878464020  -224.182666735126  -224.182666705548
7   -224.182666735803  -224.160425516424  -224.160425517110
8   -224.160425514624  -224.138610351221  -224.138610346631
9   -224.138610351305  -224.089241000935  -224.115300786446

It seems that the fifth state is an “extra” singlet state. It turns out that this state is a sextet as found by PYSCF (E = -224.295459326146 E(CI) = -12.2576489903539 S^2 = 5.9999998).

Here is the Psi4 input I use.

molecule imidazole {
0 1
  H   0.000000   2.119822   0.714354
  H   0.000000   1.202262  -1.904898
  H   0.000000  -2.104815   0.663782
  H   0.000000  -0.010302   2.116597
  C   0.000000   1.120107   0.305897
  C   0.000000   0.635508  -0.983749
  C   0.000000  -1.091835   0.283881
  N   0.000000  -0.741378  -0.994001
  N   0.000000   0.000000   1.104571
}

set {
  reference           rhf
  scf_type            pk
  basis               def2-svp
  e_convergence       10
  d_convergence       8
  maxiter             100
  docc                [15,3]
  restricted_docc     [14,0]
  active              [1,6]
  reference_sym       0
  num_roots           10
}
Ecas, wfn = energy('detci', return_wfn=True)

A question for @sherrill