In the first case the FCN is the log - likelihood built using Poisson probabilities of each bin and it is stripped by the constant term (not depending on the function parameters)
In the second case the FCN is the sum square of the residuals of each bin, normalized by the bin error