Everything Micay said in that linked thread was and remains correct. You again fail to address what was incorrect in his comment. Going on to later ask people "what is correct about it?" is rhetorically disingenuous at best.
But as you consistently slide any adjacent topic you can into a discussion about the Librem 5 (no matter how tortured a segue), let's go with that and revisit it.
I looked at your puri.sm link, and it mostly served to lower my estimation of the Librem 5's kill switch system. You can't disable the sensors in a trustworthy way without disengaging every kill switch at the same time, entering it into their Lockdown Mode. At that point it's just a still insufficiently air-gapped, highly underpowered Linux device which remains poorly secured against other side-channel attacks. The speaker which, by everything I could find, is still functional, the OS remains poorly secured against software attacks, it lacks proper hardware security, and so on.
It fails in terms of human factors, too. Joe Consumer thinks flipping off the mic switch prevents audio recording, but it doesn't in multiple regards. Even putting it into Lockdown Mode doesn't disable the speaker, which can be used to record audio despite your insistence that the device is fully secured when all switches off. Speakers can also be used to exfil data over short distances, demonstrated to work through walls.
Poor misinformed Joe Consumer is also still left with the same issues the other commenter has already identified in terms of the difficulty of securing any Linux computer.
But that's okay, because you only run trusted software. Until one of those trusted pieces of software include a compromised library, which happens often. You are, at that point, relying on the OS and its relationship to its hardware, which, flawed switch system aside, is highly insufficient. The device offers very little protection at that point. You know all this because you run Qubes OS, but hand-wave that away by appealing to trusted software as soon as the Librem 5 becomes the subject.
If I was modeling threats around protecting sensitive files on the device, not falling victim to attacks that could record audio and/or exfil data or otherwise leak, I'd still go with GrapheneOS on a Pixel 8 or later.
The Librem 5 wins for anyone who just wants a phone which runs Linux (which is a great thing and I wish we had more options which did that), but the security theater of that device is just goofy from top to bottom, as are its more vocal and less reasoned supporters. If one's threat model is, one sometimes wants to be able to turn off all radios and sensors, leaving the speaker functioning, with an otherwise poorly secured device, then, great. It's the device for you. But it's a threat model which will be practically beneficial to very few people, if any.
If your holy grail is having the radios off without other hardware or software considerations, great, you've found the phone for you. It's a brilliantly marketed device for well meaning but poorly informed people with underdeveloped threat models, and, I guess, for someone in your situation who's happy to make all of the above compromises to be able to physically disconnect radios.
Do you always enter Lockdown Mode before typing anything sensitive, due to the attack vector they highlighted about deriving typed data via sensor data? ('No, because I only run trusted software.' See above.) You literally can't disable the sensors without disabling all radios. They acknowledge that sensors are an attack vector worth addressing, yet don't put sensors on a discrete circuit. Like I said, great marketing. Otherwise pretty goofy.
Would I complain if the upcoming Motorola GrapheneOS phone had physical hardware switches? Sure, I'd take an additional layer of containment if all of the fundamentals are addressed properly.
But your argument is like bolting the world's best seat belts onto a motorcycle, and never missing an opportunity to tell the world about your belts, wonderful though they truly are.