Paul Price commented in PIPE2D-1132:

The quartz spectra in the fiberProfiles have 3x higher flux on one side of the PFI than the other, so we need to fix that or it will adversely affect sky subtraction and flux calibration. If there isn't another set with more suitable illumination, I'm hoping we can manually scale them to match the actual throughput from the sky (via a twilight exposure, maybe?).

Robert Lupton commented in same ticket:

We have to sort out the fibre throughputs. Twilight sky is one option (with field rotation!) but it's a bit tricky to get. Another option is to use the bright sky lines (e.g. 5577 in the blue, and reasonably-isolated sets of Meinel lines in the red). They should be uniform across the field of view to a percent or two, and we can beat that down as sqrt(N). It'd be good to confirm that both approaches give the same result.