Just to clarify what "
conservative" means here.
Oversimplifying, if you distribute the radiation that was released evenly in a group of 8000 persons, each one has a 50% chance of dying, so you expect to see 4000 deaths.
If you distribute the same radiation evenly in a group of 80000 persons, each one has a 5% chance of dying, so you expect to see 4000 deaths.
If you distribute the same radiation evenly in a group of 8000000 persons, each one has a 0.05% chance of dying, so you expect to see 4000 death.
But in the 8000000 group the radiation level is too low and there is also normal background radiation, and other normal risk, so the real number is probably less than 4000 deaths.
In a realistic scenario, some people get more radiation and some less radiation, but the nice part of the simplified model is that it always predict 4000 deaths. But this number is like an upper bound, the real number of deaths is probably less.