TFractionFitter strange fit

Hi, rooters,

I fit ‘data’ with TFractionFittter using three MC source distributions, all four distributions are centered at zero. As a result I get 100% fraction of my second MC source, but the distribution that I get with GetPlot() is shifted with respect to the original MC one and has a negative mean value. In the attached plot all MC distributions are in color, data are points and the fit result is in gray. Probably it is not possible in this case to fit data well, but the real question is why the fit result which is supposed to be 100% “mc1” distribution is shifted so much to the left (while the original “mc1” distribution is centered at zero), how is it possible? The other question is why the error on the fraction is so large, 65%, it often happens even when the fit succeeds for the cases when fractions are equal to 0% or to 100%. And the last question is why when setting constraints on fractions the numbering of fractions starts from 1 (i.e. I have fractions 1,2,3), but when I get results with GetResult() the numbering starts from 0 (i.e. I have fractions 0,1,2). I attach data file and script reporducing the problem. The version of ROOT is 5.30/02 (tags/v5-30-02@40973, Sep 22 2011, 10:55:04 on linuxx8664gcc). Thanks a lot for your help!

test1.eps (17.5 KB)
Test.C (1.56 KB)
test1.root (4.54 KB)