On large deviations in the averaging principle for SDE's with a "full dependence,'' revisited