The performance of a series of density functionals when tested on the prediction of the phosphane substitution energy of transition metal complexes is evaluated. The complexes FeBDA and RuCOD (BDA=benzylideneacetone, COD=cyclooctadiene) serve as reference systems, and calculated values are compared with the experimental values in THF as obtained from calorimetry. Results clearly indicate that functionals specifically developed to include dispersion interactions usually outperform other functionals when BDA or COD substitution is considered. However, when phosphanes of different sizes are compared, functionals including dispersion interactions, at odd with experimental evidence, predict that larger phosphanes bind more strongly than smaller phosphanes, while functionals not including dispersion interaction reproduce the experimental trends with reasonable accuracy. In case of the DFT-D functionals, inclusion of a cut-off distance on the dispersive term resolves this issue, and results in a rather robust behavior whatever ligand substitution reaction is considered.