Quote Originally Posted by PeterB View Post
As you mentioned I should be comparing the middle curve to one below it, but the effect is too small to observe with the limited data present in those curves. If there was more data in the curves showing shoulders for each dev time then it would be easy to compare two middle curves, but in lieu of that missing data it is not unreasonable to assume the effect we see at the longest dev time exists in the middle curves at their shoulders if sufficient over exposure occurs.
You cannot assume that. Development to higher contrasts can distort the upper end of the curve.

If you compare the two bottom curves there is no obvious shouldering in either curve. Any blips or dips or kinks are insignificant. The curves are meant to be representative averages. Depending on batch, storage conditions etc. you don't know what exact curve you're dealing with anyway.

You cannot gain information by reducing development vs an approximately normal contrast index.

Develop normally, or to N-1 if you prefer (we don't even know what your personal "N" contrast index is), and go forth.