Interesting and thank you. Fortunately the absence of GPRS or otherwise for the Binge WhatsApp exchange wouldn't affect the rest of the GPRS argument. I had it as the target for the potentially 'tampered with' WhatsApps from 22:30-01:48, but if the phone has simply been wiped of all WhatsApps there isn't a need for this (although it would make an even more cunning subterfuge as, if the messages could be recovered somehow, they would still support the evidence).
But now for your points. As I understand Moller's chart, GPRS activations are only shown where they start, they are never a continuation of something that started earlier. So the 2G at 20:50-21:00 has to start no earlier than the end of the Binge call, e.g. 20:54:04. This is an example of me placing the time based on supporting factors. I'm also being generous because the existing GPRS connection would have to cut out first and it always lasts longer than the call (ref other calls), even if OP is achieving this manually (it takes a little time). I have changed the earliest start time in the range to 20:54:04 to reflect this point (and the subsequent earliest start to be 13 seconds later).
I have the GPRS connection covering the call being 1,755 seconds which means the earliest it can start is 20:24:49 if it ends at the same time as the call (unlikely) but just 18 seconds before the call starts. This has to cover the ringing time too!
However, I can't read the duration clearly and, as I have noted, it could be 1,795 seconds, adding a further 40 seconds to the 18, making a 58 second maximum pre-call connection. I don't buy it. Binge is clear that the exchange starts at 20:10 (not 'just prior' or similar) and both he and OP talk about a conversation going back and fore before OP calls.
Also note that there was a known bug in iOS 6 around this time whereby the signal could drop not long after a call ended. Perhaps we are seeing this (check the call to Jenna). 2G cuts in first, then 3G.
Finally, Reeva's switch to 2G at 20:04:00. I think this is simply her putting the phone down by the kettle! or perhaps walking upstairs past all the metal railings.
ETA1: After further thought I see I may have missed your key point re 2G and voice but not data (which is correct). Let me think about this and come back in a mo.
ETA2: After more further thought. Your theory makes a lot of sense but it is not what Moller's chart or Vodacom are reporting so I don't think your theory works. We know from Moller's charts that when a call is placed and the cell tower changes this is reported as one continuous GPRS connection on the original cell tower, with a duration spanning the call. The new cell tower is noted against the call (and if this changes mid-call it would not be recorded anywhere). I show the cell tower change merely by adding it in the list (examples are 17:36:27, 17:44:35 and 17:53:24). Note that in all instances the last call is over 2G but the GPRS duration spans all calls, irrespective of 3G or 2G. This is also the case for the Binge call. The duration is too similar to the length of the call duration to be anything else. Vodacom must add data and voice together to create a pseudo GPRS connection time. It's more like saying it is the signal duration from the time GPRS is initiated with the cell tower through to the earliest of GPRS dropping or the call ending time.
