We investigate the suitability of density functional theory (DFT) and second order Moller-Plesset perturbation theory (MP2) for the title reaction, which serves as a model to represent the key step in a recently developed B-C bond formation reaction. CBS-QB3 is employed as a reference throughout this study. The classical barrier height associated with the concerted transition state for the H/Br exchange reaction poses a serious challenge to most standard GGAs or hybrid functionals. In particular the popular B3LYP hybrid functional shows dramatically overestimated reaction barriers (by 12 kcal mol(-1)) for the largest system with R = C2H5. We find that a proper description of intramolecular dispersion interactions arising in the transition state is crucial for a correct assessment of this reaction and that the inclusion of Grimme's empirical dispersion correction effectively compensates for most of the errors to a large extent. In conclusion we find a pleasing performance of the dispersion corrected functionals B2PLYP-D or B3LYP-D for the present set of systems if used in combination with basis sets of triple-zeta quality, which we recommend for future quantum chemical studies on related systems. Also the recently devised M05-2X hybrid meta-GGA shows an excellent performance, in particular if used in combination with the small SVP basis.