CP-MCSCF bug

[Chris Evenhuis]

I am calculating the the energy derivatives and NACMEs for H3 using 
6-311G.  I am using two state averaging but have been looking at the 
other states by using
	state,3 weight,1,1,0
and
	state,8 weight,1,1,0,0,0,0,0,0

For some geometries there is a large difference between the NACMEs and 
the difference of the energy gradients.  However, the energies, the sum 
of energy gradients and the CI vectors are equal.  The  discrepancy in 
the NACME falls of inversely as the geometry is distorted away from the 
bad point so it looks like a conical intersection (all energies are 
well separated though).

the input files are

  ***,h3
  geomtyp=xyz
  gprint,orbitals,civector
  geomtyp=xyz
  geometry={nosym;noorient;
  3
  a comment
   H1, 0.0000000000000000E+00, 0.0000000000000000E+00, 
0.0000000000000000E+00
   H2, 0.0000000000000000E+00, 0.1073111972324246E+01, 
0.0000000000000000E+00
   H3, 0.0000000000000000E+00,-0.8305234896369812E+00, 
0.4101461803747558E+01
  }
  basis=6-311G
  hf,3,1,0
     multi
      wf,3,1,1
!!!!!  Uncomment one or other !!!!
!     state,8
!     weight,1,1,0,0,0,0,0,0
!     state,3
!     weight,1,1,0
      start,3003.2
      accuracy,gradient=1.0e-8,step=1.0e-4,energy=1.0e-8
      noextra
      maxiter, 40;
      iterations;dont,internal,1,5;end;
      iterations;uncouple,5,to,10;end;
      orbitals,3000.2

      cpmcscf, grad,1.1,spin=0.5,accu=1.d-11,save=5101.1
      cpmcscf, grad,2.1,spin=0.5,accu=1.d-11,save=5102.1
      cpmcscf, dgrad,1.1,2.1,    accu=1.d-11,save=5103.1
      cpmcscf, nacm, 1.1,2.1,    accu=1.d-11,save=5104.1

  forces;samc,5101.1;
  forces;samc,5102.1;
  forces;samc,5103.1;
  forces;samc,5104.1;