I haven't dug into this data, but there's a big risk of baking in survivorship bias. In vaccinated populations, many more vulnerable individuals are still alive because of vaccination. In unvaccinated (natural immunity groups), you've culled the most susceptible individuals already (they're dead).
Think of dumbed down extreme version like this. Imagine 1% of the pop will die if they catch the virus due to a particular risk-factor. The other 99% will always be fine no matter what, but no one knows if they are at risk. For that 1%, it's 90% fatal if your unvaxed, 10% fatal if you are vaxed. What happens over time?
Vaccination rates are ~70%, so let's assume that's how our sample of 1 million breaks down and that everyone is exposed/infected.
After Wave 1:
*300k unvaxed, 1% susceptible, 90% fatal for them = 2,700 dead. But most importantly, only 300 susceptible individuals left alive. Overall IFR of 0.9%
*700k vaxed, 1% susceptible, 10% fatal for them = 700 deaths, but most importantly, 6,300 susceptible still alive. IFR of 0.1% (8x better than unvaxxed)
Wave 2 hits the now culled population
*300 unvaxed susceptible individuals alive, 90% fatal = 270 death, 30 still alive. IFR in Wave 2 of 0.09%. Wow a 10x improvement!
*6,300 vaxed susceptible individuals alive, 10% fatal = 630 deaths, 5,670 still alive. IFR in Wave 2 of 0.09%.
Wave 3 Hits further culled population
*30 unvaxed susceptible individuals alive, 90% fatal, 27 deaths, 3 still alive. IFR in Wave 3 down to 0.009%. Another 10x improvement!
*Vaxxed pop has 567 deaths and an IFR of 0.08%.
So by Wave 3, the naturally immune (unvaxed) group has an IFR about 10x better than the vaxed group. The reality of course is that the vaccinated group is still protecting their vulnerable members at a much higher rate, the naturally immune group just doesn't have many people left who are susceptible.
You can see how easy it would be to use this data as evidence that natural immunity is providing better and better protection over time. I'm not saying that's what these studies are showing, but it would be really easy for people with a narrative to ignore the asterisks/qualifications in these assessments and publish "news stories" with erroneous narratives.