I don’t disagree with what you say Robin. If you have a technique that delivers the accuracy you need for your objectives e.g. you SN matching work then absolutely fine. The difficulty I have is when corrected spectra or even flux calibrated spectra are presented with no estimate of their errors in the continuum level. (I don’t think the BeSS standard requires it.)
If you are using your own data I am a great believer in doing the minimum processing consistent with your goals to minimise added noise and potential artifacts. As we have commented before it is different when adding data to a general use database.