E.g. a 45-year-old with a latent colorectal cancer who would previously not have been diagnosed early, but only late when they developed symtpoms, by which time they hit 50, would have counted as an incidence or a likely fatality, among the 50+ data. But if that same individual had been caught at 45, they would have counted as an incidence against int he under-50 cohort.
Earlier, better and more available screening alone will shift the data this way.