In continuation of part one [J. Chem. Phys. 107, 10242 (1997)] we further examine the sources of ligand-ligand correlations in binding of proteins to DNA. We present four reasons for expecting that, in contrast to the ligand-ligand interactions, the correlations between ligands in these systems can be long ranged and nonadditive. In view of these findings we reexamine a specific set of experimental data on the binding of lambda repressor to the corresponding operators.