So I've had a look at the tar1090 and associated code (readsb for example), which has served to remind me a little of my last interaction with the software.
I hasten to say that my past experience and current appreciation of the work put into this open-source software is entirely positive, however I don't think I'd use it for the purposes of accurately determining what happened in the last few moments of this sad event.
For example; while my perusal of the code was necessarily cursory I saw several places where there is some apparent massaging of the speed component, depending upon certain parameters. Thus without detailed examination (along with the raw unfiltered ads-b out data) I'd treat the presentation you've seen as of 'interest' rather than something of serious investigatory value.
Furthermore, while it seems that raw data may be available upon request from ADS-B Exchange I assume at this point that it's still that derived from readsb or similar, and thus may contain some interpretive component(s). Although that's probably perfectly reasonable for most purposes it could muddy the water in extreme events such as this.
That said if someone with good experience of the code came along and showed me an error in my assumptions, and/or how to account for the filtering/extract the most accurate detail from the resultant datastream, I'd be happy to acknowledge that.
Coming to your comments about having a G-meter; do you utilise a purpose-built device (or maybe these days your telephone with some suitable application)? I have a personal interest in asking because I once developed an inexpensive unit to record, plot, and transmit specific seismic events - it could be easily adapted for aviation use if that was at all useful...
Otherwise thanks for the suggestion, but I can assure you I've not thought about spinning a Caravan! My last experience was in a 172, and a 150, and for me that was enough ;-)
FP.