I think a better graph would be a graph that nearly looks the same, but with the 'narrowband, wideband etc.' sections removed and the color of the line determines the bandwidth.
Also, keep in mind that this figure is meant to give a rough idea of the quality. One shouldn't read too much out of it, especially considering that we're mixing in speech, music, mono stereo, and 5 different bandwidths.
What if instead drawing fictional lines we would organize public listening test? Previous tests were done at 64 and 96 kbps. So possibilities are 128k or 32-48k test. All 3 codecs Vorbis/AAC/Opus would be on par at 128 kbps with an excellent scores about ~4.7-4.8 approaching transparency. It might be interesting but in pretty limited cases if You ask me. So we end with ~32-48 kbps. HE-AACv2,v1 and Opus are the only serious competitors at those rates (Vorbis and MP3 won't be competitive at such low rates. still can be included for a full picture).Opus 1.2.1 already does very good job and it's to suppose that 1.3 will be even better at those rates.It's a possibility for low rate test in 2018. What do you think?
Support for all of the fixes in draft-ietf-codec-opus-update-06 (the mono downmix and the folding fixes need --enable-update-draft)
MP3 is the worst lossy codec ever released, so move to Opus ;