Yes the referenced RV value for Vega in SIMBAD is very strange. I think it might actually be the rotational velocity (vsini). Even Wikipedia has a better reference for RV
Yes the transfer function is different for lamp and star. The lamp was mounted at the edge of the aperture aligned with the slit and the lines are very narrow with no aberrations compared with the star. There will always be some systematic offsets so to measure absolute RV, you need to compare with an RV standard . (In this case I was only interested in the variations in Deneb, not the absolute value so the points plotted on the graph are the uncorrected measured values) Attached is a photo showing the lamp mount. The lamp is on a pivot and moved in front of the aperture when needed.
This was a very bright target so you would probably need to reduce the intensity of the lamp (perhaps using a diffuser) for longer exposures on fainter targets. (A smaller lamp would be better, particularly for your small aperture!). The lamp lines also contaminate the spectrum but these can removed with background subtraction if a clean spectrum is needed.
The stability of the spectrograph is not the only problem with slit spectrographs. Changes in focus and guiding can also affect the way the star image is sampled by the slit. If this is not symmetric this can produce significant errors. This is another reason why fibre fed spectrographs are better for high RV accuracy.
I have also used telluric lines successfully to give high RV precision. Here in fig 6 of this paper for example.
The O2 telluric lines used in this case were very strong though.
Filly_lamp_mounting.jpg [ 106.6 KiB | Viewed 704 times ]