mm43, by the time they changed over their selcal to Africa they were effectively out of range of VHF communications, as I recall. Let's see, it was about 350 miles from the mainland and 200+ miles to Ilha Fernando de Noranha, the closest sizeable island. The mainland was completely out of range. The island was 140 miles away and VHF range from a 35,000' aircraft and a 200' tower is about 240 miles. So at their speed they had, at the crash site, about 15 to 20 minutes worth of increasingly poor VHF communications and HF communications that were extremely noisy with lightning static (QRN). Based on some short wave listening I've done on aircraft frequencies they do not SEEM to work out to the extreme VHF ranges. So the switch over makes sense EXCEPT that they did not assure their HF link was working. That was bad communications discipline on their part or a very poor set of standards they were following. Of course, I tend to assume "it's going to break when it will hurt me a lot." They followed a pattern supported by an assumption that it's going to work so no communications is no big deal. Usually that attitude is good. I prefer an "always" philosophy, myself.
"Bear in mind that ATLANTICO and DAKAR don't share the same HF suite of frequencies,..." Of course not. Otherwise they'd spend the day talking to each other or at least interfering with each other. {^_-}
I can understand switching to DAKAR, calling, getting no response, presuming they are busy, and waiting a few minutes to try again. After the second time "a few minutes" becomes one minute. After any third try I'd be asking nearby aircraft, on VHF, to relay for me. Maybe that attitude is why ham radio operators tend to get disaster communications through when everything else fails if there is a ham radio operator around.
"Remember the failed auto connect on ADS-C doesn't necessarily indicate cockpit activity or not." And that observation agrees with my comment that the event tree needs to step back at least that far when creating branches. I don't believe any of the ACARS messages really indicate an alive and aware cockpit crew, either. Although I'd suggest the communications break followed by recovery MIGHT indicate a severe attitude change and recovery indicative of somebody at the controls.
The CVR might have enough recording on it to probe backwards on this.
As a comment, it may be that the ADS-C carries automation a little too far, at least for my tastes. And usual (if not standard) cockpit practices don't put any form of premium on being in communications at all times possible. (That might even mean an Iridium phone in the cockpit over some parts of the Pacific Ocean and Antarctic routes. Color me paranoid. When things pickle I don't want to have to fumble around setting frequencies and hoping I have communications. It'd be nice to know as many of the remains of the PX would be recovered for repatriation as possible. Heck, in this case active communications might have allowed a wider detour around the bad weather spot leading to no crash at all.)
{^_^} (As a side observation note that the modern US military seems to operate on a "communicate first then act" basis now that they have very ubiquitous and nearly pain free communications at "amazing" data rates even on HF. As an engineer with over 50 years of experience in communications that in itself is a reward for my career.)