You're conflating trends within aggregated data and individual data points. Moving towards a healthier population would reduce cancer rates across the total population. It would not remove cancer entirely. Likewise any arbitrary person might be in complete health & get cancer and vice versa as you say. The two are not contradictory.
A similar example is with early detection. The data show that at population scale a lot of early detection efforts cause more harm than good. The problem is that any arbitrary person might have their life saved by said early detection. It's impossible to know who those arbitrary people will be.
A similar example is with early detection. The data show that at population scale a lot of early detection efforts cause more harm than good. The problem is that any arbitrary person might have their life saved by said early detection. It's impossible to know who those arbitrary people will be.