Metal–organic frameworks (MOFs) offer a promising architecture for designing heterogeneous catalysts, in which catalytically active metal centers can be installed on a porous MOF structure and modified using nonstructural ligands. Recently, it was shown that the activity for ethylene hydrogenation of Ni deposited on the MOF NU-1000 can be amplified 26-fold via the use of substituted nonstructural benzoate ligands (Liu et al. ACS Catal. 2019, 9, 3198–3207). However, as profound as this result is experimentally, this change in activity corresponds to an activation energy difference of only about 2.4 kcal/mol, putting it at the edge of the accuracy of density functional theory. In this study, we evaluate various density functionals for their ability to provide a consistent description of the experimental trend using a single-metal approximation of the active site. Our findings demonstrate the difficulties faced when calculating trends in small energy differences in MOFs, such as variance between functionals and differences in predictions due to subtle geometrical or rovibrational effects. In summary, we find no agreement in even the sign of the trend between different functionals, and only a minority of functionals are able to reproduce the experimental trend. However, among the functionals we examined, a subset demonstrate robustness in their predictions (either in agreement with or against the experimental trend) and we recommend using them for the computation of these small energy differences in MOF catalysis.