As I understand it, Dupuy wanted a set of broad guidelines for rating weapons that would allow the rest of the QJM to have a structure. They haven't really been taken apart and analysed factor by factor against a large body of data to determine how exactly and in what proportion each relates to casualties or combat outcome, whether all their proportions are just right, a process that involves quite a lot of time, and access to a lot of data that is often hard to find (sometimes because there is none yet). And a process I suspect Dupuy would have regarded as more trouble than such demanding work was worth. He wanted a broadly applicable model, with reasonably rough tolerances, if you will, they he could put often aggregate and estimated data into the formula to get a general answer, which could then be revised with reasoned inferences. Since battles are chaotic affairs, highly variable and poorly recorded, the more precise the model, the tighter it's tolerances, the less battle data it will accept without giving absurd results from time to time.