An alternative way of dealing with this type of data would actually be to treat it as if it were Oke & Gunn data and use `cspike' and `interp' as described in the section on the second step for Oke & Gunn data). This is probably satisfactory so long as the data does not cover a range with significant absorption features. (If it does, you can always remove the bad points with `spied' before interpolating.)
If you do this, you have to supply a bandwidth, since the table file will not have specified one. You can put one into the spiketrum generated by `gspike' (section on first step).
ICL> setobj value=40 object=gmspike.MORE.FIGARO.TABLE.BANDWIDTH
will set the bandwidth to 40 Angstroms. The effect of this is going to be similar to smoothing the observed spectrum with a filter 40 Angstroms wide in order to get the continuum values at the points specified by the table. After that, you just go through the `cspike', (`spied'), `interp' sequence as in the section on the second step for Oke & Gunn data, finally ending up with a calibration spectrum called `calib'.
FIGARO A general data reduction system