To reproduce the theoretical results published in literature, it is not a easy thing because sometimes a simulation parameter or step that seems trivial could also give surprising results. First, one had better adopt the same parameters and geometry structure with those given in literature as possible. If these calculation setups were not listed completely in literature, usually one had better try more cases. If the worst thing happens, the best way is to communicate with the corresponding author of literature.
Did you try different basis set and size of k-point sampling?
