Solvent effect on activated rate processes: On the validity of the GLE approach