The potentially pathological example of LaO has been chosen to test the application of ligand field theory (LFT) to metal monoxides. The test consists of a comparison of closed-shell ligand LFT (CSLLFT) results (in which a 2(+)/2(-) ionic M(2+)O(2-) model is a priori postulated with a point-charge ligand) against ab initio multiconfiguration self-consistent-field-multireference configuration interaction (MCSCF-MRCI) results (in which no ionicity is assumed a priori and an internal structure is allowed for the ligand). Special care has been devoted to the determination of a La3+ pseudopotential and its associated atomic basis set in order to keep the ab initio model close to the LFT one, yet at the same time capable of clearly exhibiting the consequences and the importance of the restrictions imposed in the CSLLFT model. The ab initio calculations reveal that the effective (Mulliken) ionicity in LaO is not La2+O2- but quite close to La+O-. Despite this, the (2(+)/2(-)) ionic CSLLFT model leads to the correct orbital occupations in the ground state but this model cannot account for the significant covalency contribution via the nominal 2p sigma and 2p pi oxygen orbitals. The CSLLFT calculations correctly reproduce the excitation energies of the four lowest-lying observed states of LaO : these states are calculated to within 2000 cm(-1) and globally better than via ab initio calculations. However, the (2(+)/2(-)) ionic CSLLFT model is shown to ignore the existence of a manifold of low-lying quartet (and doublet) states of (1(+)/1(-)) ionicity that might be relevant for interpreting features of the spectrum. This result exemplifies the need for developing, beyond the first attempts made in this direction, a computationally manageable open-shell ligand theory for the frequently encountered case of predominantly ionic structures with an open-shell ligand.