It is slightly disingenius to suggest there is no evidence for the linear no-threshold model. All the epidemiological data at higher doses supports it. Low dose data is weak, we know that and it is a problem.
That said, it is clear from cell studies and animal studies that the type of radiation, dose rates, fractionation and the possible priming of repair mechanisms all play a part in low dose risk. It is reasonable to assume that low dose radiation risk is non-linear, but it is damn near impossible to actually plot the relationship.
I know there a strong proponents of the idea of a threshold. That's where this 100mSv number keeps coming from. Do keep in mind that there is a lot of money in cleaning up waste etc, and a lot of people with vested interests.
While the ICRP support the LNT as the best practical model, I'll keep using it. When they pick a better one, I (and my colleagues) will move to that.