Sounds like your problem is the signal to noise ratio in your field recordings.
Normalization is the process of scanning your file for the highest value (peak or RMS, depending on your selection) and assigning it to the new value that you have determined. In turn, the same increase is applied to the rest of your file. So for example, if you have a door slam that peaks at -10dB, and you normalize that file to -2.0dB, every sound in that file will increase by 8dB, including your noise floor. Not such a massive deal breaker if your peak is at -10dB and your noise floor lies down in the -60's, -50's. But if your peak is hovering around -30, -20? That's a big jump in perceivable difference, both in signal and noise.
As a concept, I understand Ric's position. I have suffered from being blasted by gunshots after auditioning quiet countryside winds. But in practice, I disagree with his stance. I don't want my gunshots to be as quiet as a countryside wind, nor do I want my wind as loud as a gunshot.
It's been a while since I read my "Bible" but in re-reading your post and thinking about the logic behind it, perhaps his intent was to normalize similar sounds from an individual recording session. So, for example, if I do 3 sessions of a blender, a toaster, and handling flatware, I could see the wisdom in normalizing all my blender recordings similarly, all my toaster recordings similarly, all my flatware recordings similarly... but I digress.
The main concept (one that we're all constantly chasing) is to have sufficient enough recording levels and mic placement so that you don't need to normalize your recordings after the fact.