You are correct about two sites in the same spot, and that averaging might not be right.   I neglected to mention that in our algorithm, we use site area weighting in the IDW calculation.    We we actually do an inverse weighting of area*yield, instead of yield.   But that doesn't address that you might have a valid yield of zero because the site had already been cut.   This is a real problem on wheat corners.    We've been trying to do an algorithm to detect that, but no luck, yet.

You are also correct that all of the raw data to do distance, width, flow, etc is in the source files.     Knock yourself out with it :)