Yes, you should include QE correction for your flats, standard star and science data. The “refimages” parameter should be provided with the corresponding extracted arc lamp exposure(s) (beginning with an “e”). These arc(s) should have been wavelength calibrated beforehand (with wavelength calibration and aperture trace entries still present in the database subdirectory). You can run gqecorr directly or via gfreduce.
You can find an example from a tutorial at a recent meeting at http://jastro.org/GMOS_IFU_2017 (this URL might change at some point).
Hope that helps,
James.