Wannier calculations require a full k-point set
To my understanding the wannier code requires a full uniform k-point set (i.e. 8x8x8 = 512 kpts), other wise you will get:
Error message: mysterious kpoints Error occurred in subroutine: wann_get_mp Error from PE:0/16
I am not very deep into the wannier code, but I am a little surprised by this, because in other parts of fleur we can reconstruct the missing eigenvectors through the symmetry trafos. We could check if SPEX uses the symmetries for their wannier generation.
Maybe it is possible to extend this to the wannier part. Which would be quite important, especially for larger calculations.