Large Maximum Im/Re Ratio

[GCatarina]

Wannier90 for benzene is giving me sensible WFs (C:pz) and Hamiltonian, but I keep getting large "Im/Re ratios" (~100-1000), which from what I read suggest that I am getting stuck in a local minimum.
Input and output files below.

I am using QE (tried both v7.2 and v7.5), and I already tried both the stable v3.1 and the developer version of wannier90, with similar results.

For completeness, I also tried this example for methane and I still obtain large Im/Re ratios.
I will now explore the random noise functionality (conv_noise_*), which I never tried.
Any other suggestion?

wan.win.txt
wan.wout.txt

[AndresOrtegaGuerrero]

@giovannipizzi Do you know what could be the reason of this? , We also tested the library mode with the PySCF interface and we get zero for this ratio, we wonder the source of the difference

[GCatarina]

Update: exact same calculation but without the 'gamma_only' option (both in pw.x and wannier90) gives similar outputs (same spreads, WFs and Hamiltonian also look the same) but now the Im/Re ratio is very close to zero