Phase calibration is the key operation of phase closure imaging. In the case of nonredundant arrays, the related problem amounts to finding the node of a Z lattice closest to the end of a vector, the components of which are the differences between the closure phases of the data and those of the model. The aim of the paper is to show that this integer ambiguity problem can be solved in a very efficient manner. Its potential instabilities can also be well identified. The corresponding approach, which can be extended to redundant arrays, revisits and completes the analysis presented in a recent paper entitled "Phase calibration on interferometric graphs" [J. Opt. Soc. Am. A 16, 443 (1999)]. The procedures of self-calibration must be modified accordingly. The spinoffs of this approach also concern the integer ambiguity problems encountered in the global positioning system.