That data is biased by the small minority of super active users who spend 4-5 hours/day on Twitter.
The relevant number would be to condition on only the 20th to 80th percentile users (by time spent/day) and see their breakdown. I am going to bet that number is more biased towards desktop, while both the 0-20% (occasional users) and 80-100% percentiles will be mobile focused.
The other confounding effect is the bots and the pseudo-bots (humans operating many accounts). I don't know how they change these numbers.
I don't know a single person who willingly prefers a limited, bogged-down experience of any product (social media or otherwise) over a full-fledged desktop client.
Of course I prefer doing all of my computing on my desktop computer. But I don't take my desktop computer with me to the grocery store. And I use Twitter to entertain myself when I'm waiting in line at the grocery store, for example. When I am at my desktop computer, I'm much less likely to be interested in using Twitter.
That said, one of many reasons I didn't bother creating a Threads account is precisely that reason: no desktop client. Unless forced, I won't use anything that is exclusively available on mobile devices.
I have mod rights on a general subreddit (a city), and I see that mobile is consistently 75-80% of the traffic. That seems consistent with the numbers above.
The relevant number would be to condition on only the 20th to 80th percentile users (by time spent/day) and see their breakdown. I am going to bet that number is more biased towards desktop, while both the 0-20% (occasional users) and 80-100% percentiles will be mobile focused.
The other confounding effect is the bots and the pseudo-bots (humans operating many accounts). I don't know how they change these numbers.