One issue with relative intensity spectra in that where you place the normalisation region significantly effects the plots.
In effect you force the changes to be other than at that point. I don’t think the is a simple solution but requires going to absolute flux as described by David Boyd.
Regards Andrew