Problem with fittings in a loop

I am trying to fit an array of TGraphErrors inside a loop using TF1 pol2, however this just works fine for the first fit, when I enter again in the loop the result I obtain is very similar to the first fit (which is not true in my case). How can I reset the pol2 function so that it will forget the previous fitting and will just take into account the current graph?

Another funny thing is that if I include the B option (in which I should enter the initial parameters mysef) but I don’t enter the initial parameters, I get a sensible fitting. However if I don’t write B, I get a completely crazy fit.

TGraphErrors *grfit[10];

for (i=0; i<10; i++)
{
grfit[i]=new TGraphErrors(n, x, y, errx, erry);
grfit[i]->Fit(“pol2”, “QB+”, “”, R1, R2);
}

Please send the shortest possible script reproducing the problem

Rene