A concern:
How would one go about differentiating between those potential checks which are blocked by pieces and those that are not? It seems as if this would be difficult to overcome when attempting to obtain any type of upper bound on the total number of illegal positions, given your first assumption. The other assumptions I don't believe would shave away many of the possibilities, relatively speaking, with the possible trivial assumption of no pawns on back ranks.