The spreading out effect is a direct consequence of the linear no threshold model. If you have a total dose D spread over N people, then everyone receives a dose D/N. With some constant k converting individual dose to cancer rates, you end up with an expected number of k D/N * N cases of cancer. ( So it is almost certainly wrong, however it is not understood how wrong it is.)