Allow the pre-2020 slope to carry on without an artificial discontinuity.
I figured all the action would be in the TFR projections, and the main surprise for me was that the effect of Covid is so clearly visible even though the data ends at 2020. Just eyeballing the chart, this model appears to take that dramatic downward turn in fertility and fit a curve to it alone, ignoring pre-2019 trends. This is evident in particular for Asia, a rather important contributor to world population trends.
Perhaps as climate change and other problems continue to worsen, fertility rates will start to depend more and more on unexpected events that make people scared and pessimistic about the future, in which case this projection could prove accurate.