Instead of interpolating between "known calibration values at given zoom settings" they allow for the estimation of a separate interior orientation parameter set for each image (see calibration groups) so I am not exactly sure about what you have read about the implementation of Brown's model or how you have interpreted it.
I suppose that what has been mentioned is that it is better to estimate 8 parameters instead of 8000 so it is better to keep zoom,focus,aperture fixed.
I hope this helps