BSE: convergence vs. kpoints mesh

[matdisor]

Hi,

I was using BSE to for calculating optical absorption spectra of a 2D material.
I see that convergence vs. kpoint is not that good. I get qualitatively similar spectra from k-point from 30x30x1 to higher, say 45x45. The only problem is the shift to lower energy (100 meV). I though 30x30 is quite dense mesh. The shift is fine (eventually it effects strongly on the binding energy of exciton), just want to have converged results.

Is it a common problem? Is there any setting to improve the way the code integrate over the BZ?

Thanks.
Duy