It was done rather crudely in this example. The spectra are flux calibrated using a nearby reference as normal. I then just subtracted the Pickles G2v spectrum scaled relative to the comet spectrum in a region where there is just the dust continuum. It is pretty good at removing the solar absorption lines and most of the dust continuum but leaves a residual shape in the continuum dependent on the broad reflectance spectrum of the dust so not good for quantitative measurements. It would be better to modify the G2v spectrum first to better fit the dust continuum leaving just the emission.