On the calibration of line-of-sight magnetograms