As we forced the gluon-rich fraction to 0, the chi2 value decreased. It caused the concern that maybe the fit procedure takes a local minimum. However, it is important to remember that we use a log likelihood method and therefore, we do not minimize the chi2 and so, there is no reason for the log likelihood and the chi2 to have the same minimum. We used a chi2 method instead of the log likelihood method for comparison. The resultant chi2 fit can be seen here. The chi^2 fit is an increasingly poor choice with small statistics, and so the binned max likelihood is clearly better technically and theoretically and it is what we use for the analysis.
There were some old parameterization values used to find the gluon-rich fractions and as I changed them to use the correct values, some of the fractions changed slightly. If you see a difference in the numbers from the blessing talk, is due to this fix.