but was probably actually 280 at first and was actually 320 by the end. How did you determine this?
FTP tests (I round to the nearest 5w).
Since you seem to be looking at my TR data in detail without knowing full context… this year I only really did TR consistently between maybe the end of Jan and March. Outside of that I had been doing other things and not syncing them into TR - I don’t have any events this year and am not really pushing the structured training for the sake of mental recovery after probably overdoing what was sustainable for me mentally for the past 1-2y. There is a whole bunch of data outside of those dates that isn’t uploaded to TR… because I didn’t really care that much lol.
It looks like there is a 5-month gap between your last two FTP updates and an 8-month gap between the two before that.
The 8month gap is because I wasn’t using TR at all. I was working with a coach then and using TP, but some things were still syncing into TR. There were many FTP tests and changes in FTP in that time window that are just not recorded in TR. I think my FTP in Oct 2022 was around 260 (coming off a break after riding across the Pyrenees), and it was 330-340 in July 2023.
The 5month gap is when I was doing kind of whatever I felt like that given week, and trying to figure out if I was ready to do structured training again, getting over illnesses, etc. again, not really using TR religiously or using it to track everything. Basically just futzing about and trying to keep busy until I figured out which direction to go.
There are also a lot of manual changes to your FTP that seem to go back and forth, up and down, regularly.
Not sure what timeline you’re looking at as I think I’ve only made two changes in my FTP setting in TR within the last 12 months. If you’re looking at data from ~2022-2023, then that’s all when I was using TP and working with a coach. Anything prior to 2022 I don’t remember in detail lol. But yes, in that era (2018-2022?) there was a lot of up and down and that was actually one of the reasons why I’d stopped using TR and moved to using a coach - didn’t seem to be making any meaningful progress.
removing the HR data from the equation shouldn’t make a massive difference since you have so much good power data available for reference
Maybe it was just a coincidence then. But I thought it was interesting that AIFTP kept on giving me 300w +/- 5w for the few months I hadn’t been syncing HR data into TR anymore… regardless of what my actual FTP had been recently… when in the past I’d found it pretty accurate… and it then ended up being more accurate again once I added HR data back in. 
And I don’t think Id describe the difference as “massive.” We’re talking about AIFTP going from like 305 to 315 after adding in a few days of HR data when FTP was tested at 320, which is only a 3% change.