General analysis of the consistency of nuclear data and verification of the optimal values of excited state energies, presented in the evaluated nuclear structure data file (ENSDF) (October 2005) have been performed. As optimality criteria, we chose the Pearson and Romanovsky criteria. As a result, along with finding some misprints, it was revealed that the level energies are not optimal for a large number of the ADOPTED LEVELS, GAMMAS sets included in the ENSDF. The use of the standard GTOL program (7.1a version), which calculates the optimal values of the nuclear level energies from the energies of the transitions positioned in a scheme, made it possible to significantly improve the statistical consistency of the schemes.