You can't fit that data with the sum of two Gaussians because it's clearly not the sum of two Gaussians. There is something else as well, and that has to be included in the model. The best approach would be to figure out where that "something else" is coming from, and model it based on the physics. Absent that knowledge, we have to model it with some arbitrary function. Fred added one more Gaussian. I have attached a worksheet with a somewhat more general solution. First, I don't know where that tail comes from, but it has nothing to do with the two Gaussians, and seriously affects the fit. So I truncated the data to get rid of it. Then I included a third order polynomial to model the baseline drift that is clearly present in the data.
