Abstract Supercell calculations using density functional theory with local or semi-local exchange functionals have proved to be a very successful tool in defect engineering, apparently leading to some degree of overconfidence lately. With a case study on selected defects, we demonstrate that the approximations involved in these functionals lead not only to an underestimation of the gap but also to related errors in the total energy difference of two defect configurations (be these in the same or different charge states). Sometimes the error is so serious, that even the common expectation: “DFT provides the correct ground state” is refuted. We also demonstrate, that semi-empirical hybrid exchange functionals, which reproduce the correct band gap, give good total energy differences even in these cases. In such calculations, the position of Kohn-Sham gap levels with respect to the band edges gives a good estimation of the vertical ionization energy. Based on that, we propose a simple correction scheme for checking the necessity of extended total energy calculations with a hybrid functional.