[Rivet] normalisation to cross-sections

[Frank Siegert]

Sorry, one minor correction that I realised just now sitting in the 
seminar...

Frank Siegert, Friday 23 October 2009:
> If I normalise
> histograms to 1 and I have filled _weighted_ events into these
> histograms, this doesn't work, because there might be runaway bins
> which create a bias towards low values for all other bins.

This isn't limited to _weighted_ events after all.
The main point when this appears is when a histogram is filled after very 
strict cuts, such that the integral of the histogram might be quite 
unstable from run to run.

Frank