No, I don't think you've missed anything in the regs. I should point out that the technical side isn't my forte. For purposes of this subject, I'm more into the logic of what might be if a change is/isn't made. Traditional reasoning holds that faster signals require more bandwidth. The precise relationship between bandwidth and speed has changed over time, with the development of newer modes and perhaps better transmitters...but if you fuzz up your eyes and look at it at a high level, it's still generally true. Now, consider the use of data modes on HF. In amateur use, we primarily use data for keyboard-to-keyboard communication or to move relatively small text files. Keyboard-to-keyboard modes don't need to be high speed (and yes, I'm blurring the distinction between "wpm" and symbol rate here), and effective communication can be achieved, even in weak-signal or disturbed/fading conditions with relatively little bandwidth required. There's no reason to go wider. Sending text files is a different matter. In theory, we could achieve pretty speedy throughput at a symbol rate of 300 with an ultrawide signal. However, such a transmission would be exposed to various issues with fading (although modulation choice could limit the effects, and error correction can help), and the signal would have a little less "oomph" due to the frequency span covered than a narrow signal (see, e.g., how low power CW is more effective than low power SSB for DXing). Given those issues, constrained to a 300bps symbol rate, relatively narrow modes are likely more effective than ultrawide modes. Ultrawide modes would become more logical to use only when a faster symbol rate is available -- the ability to transmit symbols faster, either to increase throughput or to better accommodate the overhead of additional error-correction overcome the headaches of fading and spreading power across additional bandwidth. Outside the ham bands, ignoring OTH radar and the like, the cool kids seem to be playing with modes that are about as wide to slightly wider than SSB voice, with symbol rates of 2000bps or more....so I suspect that it will be quite some time before a 5+ kHz wide signal makes enough sense for someone to spend time doing the development work resulting in a viable mode (ignoring experimenters' who dabble "just because"). But maybe that time won't be too far off, depending on what happens with things like SDR.