fix for problem with error calculation with weights

Remember the problem I had with BDT efficiencies > 1? This lead to events being skipped in the yield calculation, though everything was fine except the single component BDT efficiency. I now get a similar thing with the error calculation in a few bins of the BDT efficiency histos. This fixes it. If you're skeptical that this breaks the whole error calculation, you can always have a look at the cout and see if it happens at a critical step.