Parametric modulation modulates the within-subject response by the parametric value. The canonical HRF value shouldn't change as the PM is orthogonal to the canonical HRF term.
The parametric regressor term represents the slope of the effect of the parametric modulator in each subject.
It seems that you interest is in the effect between subjects of the canonical HRF, but accounting for any subject differences in frequency. As such, you'd want to account for frequency in the group model as the PM term measures the within-subject effect of frequency.