Quantifying differences between computational results and measurements in the case of a large-scale well-confined fire scenario