Mark bad channels, interpolate and CAR: interpolate before or after CAR?

What is the best practice?

I have a dataset on which I am:

  • marking the bad channels
  • adding the reference channel
  • filtering between (1., 45.) Hz
  • computing a CAR reference and adding it as a projector (not applied)
  • interpolating bad channels
  • applying the CAR projector

Obviously, I want to compute CAR after marking the bad channels. But should I interpolate the bad channels before? Does it actually matter?

I am a bit confused, and I’m not sure what is happening to the projector when interpolating the bad channels.