In such case, the omit map is an electron density map calculated using the experimental amplitudes and the phases derived from a model without some chosen atoms (your ligand), which were selected to be omitted. In the region of those omitted atoms, the map will be less "biased" and, therefore, more informative. Phenix.polder is a great tool to generate such a map.

To calculate a map the experimental amplitudes are combined with the phases from the model. This introduces a phase bias where the phases from the model can bring back electron density that isn't in the data. By omitting the ligand from your model before calculating the omit map you prove that electron density for your ligand comes from your data and not from the model.