First of all, it is well-known that the ground state of graphene is non-magnetic. so it is meaningless to do the spin-polarized calculations (e.g., LSDA, SGGA), because they will give the exact results of non-spin-polarized calculations.
Second, graphene is zero-gap, so a quite dense k-mesh for the k-point sampling in the self-consistent calculations is needed. The 9x9 k-mesh may be not sufficient.
When a supercell is used, the obtained band structure is not easy to compare with the one obtained by the primitive unit cell, because the band folding.
The difference between your results and the published one mainly exists for the sigma bands. You may check the lattice constant of graphene and the C-C bond length in your calculations and those used in the published paper. Are they same?