I'm guessing those were positive and negative fudge factors. Most cancelled out, but the end result is the footnote level corrections are larger in absolute mag

I'm guessing those were positive and negative fudge factors. Most cancelled out, but the end result is the footnote level corrections are larger in absolute magnitude than the underlying budget, making it very difficult to parse (imagine if you had so much monkey patching nothing of the original code base remained).