I am thinking... if you choose in band, create new fresh digitized DTMF to transmit as digitized audio instead, or does it just digitize the DTMF produced by the phone and pass that along. If the later, maybe raising ChannelRxGain by 6 or 12 dB would help if the tones were being mostly recognized but not always.
Anyway, the latest word from the OP is that things are working well. So this is more my curiosity at this point. I did increase my ChannelRxGain by the maximum (12 dB) to get the audio level reports by the other side of the calls to be better. I also changed ChannelTxGain to -3 dB to reduce the audio to my ear.
So I don't know if ChannelRxGain would affect the in-band DTMF or not. I offer it to those who experiment.