I think he just overcomplicated it with the bits thing. He was trying to draw a parallel in your head about the actual resolution difference ( in digital terms ) about full use of headroom vs. line level use of headroom. He did state clearly that we need the extra headroom for accidental and unknown peaks to be allowed to pass without clipping the HA.
We can sit there all day and mess with the HA setting to get the signal to be as close as possible to line level, unity gain, or whatever you want to call it. The plain simpleness of it though is that we just want to be close to that. I think when he means RMS he means that the signal, for the most part, sets right at line level and occasionally peaks come through that go above that but don't clip. without trying to play god with the setting, simply set the HA so that the incoming signals peak level is set at or just above the reference setting. Wham, that's it, your done, rock on.