Spin orbit using initial CHGCAR file from previoulsy converged spin-polarized calculation

[chelman]

Hi
I'm performing spin-orbit calculation to obtain the MAE (magnetic anisotropy energy). The manual states that it is possible to calculate the spin orbit from a converged spin polarized calculation, using the CHGCAR. So therefore I calculated for different directions. For each different direction the initial CHGCAR was spin polarized. But, for reasons that are long to explain in this mail, I did the same calculation but the initial CHGCAR was the previously calculated with spin orbit. To clarify, for the first direction (could be SAXIS=1 0 0) the initial CHGCAR was the obtained in spin polarized calculation. For the second direction (let's said SAXIS = sqr(1/2) sqr(1/2) 0) the initial CHGCAR was the one obtained for the first direction. For the third direction the initial CHGCAR was the one obtained for the second direction, and so on different directions.
What I obtain is:
in-plane angle Energy(1)(eV) Energy(2)(eV)
0 -192.96158 -192.96158
30 -192.96173 -192.96175
45 -192.96189 -192.96193
60 -192.96204 -192.96212
90 -192.96219 -192.96232
120 -192.96204 -192.96225
180 -192.96158 -192.96198
(1) initial CHGCAR from converged spin polarized calculation
(2) initial CHGCAR from the converged for the previous angle (except 0? was from spin pol. calc.)

So three main question:
1) Could somebody explain me why there is an energy difference between columns (1) and (2)?
2) Most important, Since the system has C_2 symmetry, at 180? the energy should be the same at 0?,so what happen in column (2), the energy at 0? should be the same at 180??
3) Could somebody provide me with any reference about the way spin orbit is implemented on non-collinear calculation in VASP?

Thanks in advance

[admin]

how large EDIFF in these calculations?

[chelman]

The EDIFF is 0.1E-04 for both cases.
Thanks