[Wien] Does the in1new flag average the spin up and spin down EPL and EPH values?

[Lee, Chi-Cheng]

Hello,
I am running wien version 8.3 with intel (+ mkl) compiler. 
To get a better radial wave function u(r), I am testing the -in1new flag for 
antiferromagnetic LaOFeAs. However, for the two antiferromagnetic
Fe1 and Fe2, I got different reference energies as listed below:
 
For example in case.in1,
Fe1:
 .42660   4   0      global e-param with N other choices, napw
 0    0.325     0.000 CONT 1
 1    0.411     0.000 CONT 1
 1   -3.141     0.002 STOP 1
 2    0.529     0.000 CONT 1
Fe2:
 .42660   4   0      global e-param with N other choices, napw
 0    0.310     0.000 CONT 1
 1    0.406     0.000 CONT 1
 1   -3.227     0.002 STOP 1
 2    0.500     0.000 CONT 1

After checking the epl and eph values in case.scf, I found the code didn't average the
epl (or eph) values between spin-up and spin-down channel. Since the u(r) is 
spin-independent (am I right?), why doesn't wien2k code average the values for
spin-up and spin-down channel? Then I can have a equal reference energies for Fe1 
and Fe2. I also try LDA+U with -in1new, the situation become worse. Because the
energy band from spin-up channel is much different from spin-down channel. In this 
case, -in1new add one local orbital for Fe2 but not for Fe1, which I want to 
treat them on equal footing. The new case.in1 from LDA+U is:
...
 .45460   4   0      global e-param with N other choices, napw
 0    0.324     0.000 CONT 1
 1    0.418     0.000 CONT 1
 1   -3.098     0.002 STOP 1
 2    0.521     0.000 CONT 1
 .45460   5   0      global e-param with N other choices, napw
 0    0.306     0.000 CONT 1
 1    0.421     0.000 CONT 1
 1   -3.316     0.002 STOP 1
 2    0.353     0.000 CONT 1
 2   -0.206     0.000 CONT 1
....
...
Should I average the energies by hand after each iteration and also add LO to make these two
atoms possessing the same definition or there is a simple way out?
 
Sincerely,
Jerome