Neutron detection inevitably proceeds via nuclear reactions, making nuclear cross section data of major importance in the determination of neutron detector properties in general, and efficiency determination in particular. Up to about 20 MeV, the cross section data base is fairly advanced, and the codes for, e.g., determination of scintillator properties produce results with uncertainties on the level of a few percent. Above 20 MeV, the scarcity of nuclear data results in much larger uncertanties in model calculations of detector performances. The data base above 20 MeV is meagre, but this is not the only problem. It also contains severe discrepancies for the most important cross sections. In this energy domain, the neutron-proton (np) scattering cross section is used as primary standard. The np scattering database contains discrepancies up to about 10 %. Since essentially all other cross sections are measured relative to np scattering, this means that most data have similar uncertainties. In the talk, the present status of reference cross sections in general and np scattering in particular is outlined, and recent experiments to elucidate the problems are presented. Moreover, the data situation concerning fast-neutron fission as reference is discussed.