It's certainly going to take a while to compile the data we'll need but I think we're going to find that recent mining of blocks far outweighs earlier mining in terms of the power requirement.  My hope is that it will be enough of a difference to make the error very small even with the assumptions that will have to be made.