LDA+U: shift by delta_E > U and not U/2

Dear exciting community,

I’m doing LDA+U calculations in exciting to shift highly localized semi-core d-states to match experimentally measured XPS energies for these states. I’m confused by the observation that these fully occupied states are not shifted by ~U/2 down in energy but rather by an energy shift being even larger than U (about 3/2*U in all my calculations, tested for several semi-core states in different materials, using the fully-localized limit).
From a brief look into the src I couldn’t find a reason but I’m wondering if it is due to the basis set or that the occupancy is treated differently such that there is a different prefactor for the energy? I tried doing a spin polarized calculation with LDA+U to compare to non-spin polarized and noticed there that the states are not at all affected by the +U, there is no spin split but the occupancy is now 1. Hence, LDA+U and spin polarized doesn’t seem to work / change the number of electrons…(or I’m doing something wrong?)

Any comments would be highly appreciated - thanks!
Nora