We study the systematic deviations of the theoretical results (the Hartree-Fock-Bogoliubov, the relativistic mean field, the Duflo-Zuker, and Weizs\"aker-Skyrme models) and experimental values of atomic masses, one-nucleon separation energies, and nuclear charge radii. Strong correlations of these deviations, which has not been discerned hitherto, are highlighted, and are applied to improve the agreement of theoretical models with experimental experimental data. Substantial refinements of the consistency between theoretical models and experimental databases are achieved with additionally only one parameter which reflects the correlation reported in this Letter. The root mean square deviation values of these theoretical models in comparison with experimental databases are about 100--300 keV for atomic masses, 120--350 keV for one-nucleon separation energies, and 0.008--0.017 fm for nuclear charge radii, for nuclei with proton number $Z$ larger than 28.