Is it possible to process the impulse excitation using the constant Q transform (CAS Journal, Nov.93, p.1-4)?

The constant Q-transform is similar to octave band smoothing of the frequency response. Please see the section "Smoothing or Integration" in "Menus and Dialog boxes.pdf" on how to use this.

The other issues described in the CAS Journal article referred to above is only valid for periodic signals, and not for an impulse response. Ideally the Q-transform is performed using time-windows of increasing size for low frequencies. This feature is not directly supported in WinMLS, but tests using time-windows of different sizes did not show significant improvement using this method.