I agree with you, like we discussed earlier in this thread. That kind of peak filter (as is shown in the paper with a few diagrams) has issues, the first is that the bandwidth of the peak is different if it’s a gain or cut, and to get different levels of boost/cut you need to apply a gain to either (or both) the LP and HP outputs. To fix that you need to use LP + (G/Q) * BP + HP, where G is the gain of the peak.
The issue there and the reason Ivan is opposed to adding it is if you want to have modulate-able peak gain and Q factor, you need to provide interpolation to smooth the gain changes (which he mentions with the LinearSmoothedVariable class). I’ve already done this myself and there are some issues with how the filter’s process loop is optimized in release builds, so it’s not exactly trivial.