The problem here is trying to devise a linear parameter to describe a nonlinear component. After all, the distance you drive depends on the route and the distance measured with calipers on a map won't agree.
So the question is, do you want to know the slope of the B/H curve at a given point and then call it inductance? Change anything, and you'll get a different number. Why do you want the number anyway? These components are just that, components, and not standards or anything else absolute. The bottom line has to do with the details of the application, and how it works depends on how you use it. That's why anything accurate requires trimming of parts, especially, say, the coils in an old fashioned organ or in an analog signal generator.
I can measure inductance accurately with a tiny signal and no bias current. I don't know if that value offers help in circuit design. Of course, an air core inductor hasn't got these problems. So for very low power applications, we will stay around the origin of the B/H curve. If there isn't enough flux to cause the core to function, you get the same thing as without the core.
I remember asking a college instructor why the curve of inductance has that first tiny bend in it, very close to the origin. I never could get an answer but now I know. What this has taught me is that, just because someone teaches a subject or writes a book on it doesn't mean he understands the subject.