Reposting the ROOT forum topic here since we did not get any feedback on it.
The attached code demonstrates incorrect calculation of RooAddPdf parameter value when fitting data with a sum of two RooAbsPdf's (exponential decays) convoluted with a RooGaussian via RooFFTConvPdf.
In order to reveal the problem we fit the same dataset by means of RooDecay's convoluted with RooGaussModel and obtain different intensity of exponential component.
Both fits are plotted on the same canvas here. Image clearly demonstrates the notable discrepancy between fractions of exponential decays. Other model parameters as well as fit curves look identical.