They also mention “people under 75”, so I assume they’re just using quarter-century ranges: 0–24; 25–49; 50–74; and 75–99.

Actually, here's a figure from the original study that uses 5-year bands, and you can see the anomaly covers those specific bands (with the under-25 bands looking more like random noise, which I assume is a statistical effect of the low number of total deaths in that range).

Why strange? This is quite a common age range where you exclude the children and college kids and go to middle age. Or you just cut a 100 year life into 4 quarters.