epsilon.x JDOS calculation
"JDOS" calculation in epsilon.x (PP) outputs unusual results for graphene. I would expect peaks at 4eV and 14eV, but the result show a wide 13eV and 25eV if "nscf" run was performed with nbnd=8, while nbnd=16 shows an even wider peak at 15eV and 26eV. Number of band should not effect the position of peaks?! Example in the attachment.
"nscf" was calculated with 60x60x1 kpoints. Different occupations tetrahedra, Gaussian and mp smearing (degauss=1e-3) give the same results. I tired both entering the kpoints manualy and automatic with nosym and noinv = .TRUE. Both Gauss and Lorentz smearing (in epsilon.x) give the same results (intrasmear=0.1eV).
On the other hand for the same "nscf" calculations "eps" gives expected results for epsr and epsi. Epsi shows peaks at 4eV and 14eV in-plane and 11.3eV and 14.5eV out-of-plane.
I checked the epsilon.x source code and I could not find a reason JDOS would be so different.