Despite the importance of vacancies over the properties of insulating oxides its influence on neighboring transition metal ions is far from being understood. This work is devoted to find the origin of various up to now unexplained properties of chromium bounded either to a < 100 > or a < 110 > Mg2+ vacancy in MgO. In these model systems particular attention is paid to understand, by means of ab initio calculations, why the cubic field splitting parameter, 10Dq, is surprisingly 1600 cm(-1) higher for a < 100 > than for a < 110 > vacancy, a fact behind the suppression of the sharp E-2 -> (4)A(2) luminescence in the latter case. Our calculations, which reproduce the main experimental facts, prove that the average Cr3+-O2- distance is the same within 0.8% for both systems, and thus, the low 10Dq value for a < 110 > vacancy is shown to be due mainly to the electrostatic potential from the missing Mg2+ ion, which increases the energy of antibonding t(2g) (similar to xy, xz, yz) levels. By contrast, for a < 100 > Mg2+ vacancy that potential provides a supplementary increase of the e(g) (similar to x(2) - y(2), 3z(2) - r(2)) level energy and thus of 10Dq. The existence of the E-2 -> (4)A(2) luminescence for Cr3+-doped MgO under perfect cubic symmetry or with a < 100 > vacancy is shown to be greatly helped by the internal electric field created by the rest of the lattice ions on the CrO69- unit, whose importance is usually ignored. The present results underline the role of ab initio calculations for unveiling the subtle effects induced by a close vacancy on the properties of transition metal ions in oxides. At the same time they stress the failure of the empirical superposition model for deriving the equilibrium geometry of C-4, and C-2, centers in MgO:Cr3+.