When Innous appeals to "noise" as a problem they've solved a good first question to ask "what noise is that?". In this case the Innuos claim seems only to make sense if it's time-domain noise [1] while Rob Watts may be writing about amplitude-domain noise [2].
Limited bandwidth over an optical digital connection plus the poor resistance of the basic SPDIF/TOSLINK digital protocol to data-dependent jitter in that situation means an optical receiver has to work hard to reject jitter (but that's been perfectly do-able for a very long time), while an asynchronous USB interface can be as perfect as the receiver's local clock jitter.
As usual, technical marketing stories from either Innuos or Chord tend to tell a convenient part of the truth, not "the whole truth". They have to be interpreted with scepticism.
[1] Jitter, wow, flutter, scrape flutter (or perhaps other names I don't know about) depending on the source.
[2] Voltage or current noise usually.