Multibanding likelihood doesn't produce matched filter SNR correctly

I'm not sure if this is an issue with Bilby or PESummary, but the matched filter SNR doesn't appear to be correct. The other SNR values (optimal etc) do look correct. This plot illustrates the SNR of a simple injection/recovery test run with IMRPhenomNSBH using different relative binning options and multibanding.

Here is the version information:

# multibanding version information

multibanding_version=bilby=2.0.1

# narrow_mc_no_seed version information

narrow_mc_no_seed_version=bilby=2.0.1

# rel_bin_inj_seed version information

rel_bin_inj_seed_version=bilby=2.0.1

# rel_bin_no_seed version information

rel_bin_no_seed_version=bilby=2.0.1

# pesummary version information

version = '0.13.10'
last_stable_release = '0.13.10'

git_hash = 'b4d82f3'
git_author = 'Charlie Hoy <charlie.hoy@ligo.org>'
git_status = 'CLEAN: All modifications committed'
git_builder = 'Charlie Hoy <charlie.hoy@port.ac.uk>'
git_build_date = '2022-10-27 14:39:44 +0000'