Which released-bracket method belongs in our FR5 comparison?

I have separated the selected old failures from ordinary trial stock; now the old FR5 proposal mixes clamped and released measurements. How should quality define the current comparison without importing the old correction as a given?