- That backdoors are routinely detectable in closed-source applications. They are not just "not easily" detectable, they are impossible to detect without a good amount of luck, and even so, the knowledge about a new backdoor will come after years, after all of the damage was done. Numerous discovered RCEs in Windows, some of which laid there for more than a decade after being found, confirm this.
- That there are incentives for security researchers to routinely disassemble WhatsApp binary and provide results of their inspection to the public, and not just report to employers or use the knowledge about a Facebook-authored backdoor to blackmail Facebook. There are none.
If you discard these two assumptions, you can see that Telegram model is better because it gives users clear choice between passing data to Telegram (in exchange for chat sync) and using E2E, with strong guarantees of E2E that are backed by reproducible builds, not by trusting Facebook or "oh, but that would be a scandal". The "almost useless" E2E of Telegram (I don't see how it's almost useless, it's there, it's working, and it provides value for users) is better than the completely useless E2E of WhatsApp.
(But yeah, group chats have no option of E2E in Telegram, that sucks.)