I think part of the issue is that the MS20 highpass is realized by swapping signal and ground in portions of a 2-stage lowpass filter, rather than swapping the R and C components around. Assuming zero resonance to simplify (so feedback = 0 V), the lowpass path is RC→RC (2nd order, 12 dB/octave) while the highpass path is a single CR (1st order, 6 dB/octave). I should probably model/measure to confirm… or get myself an MS-20, maybe.