This just shows one example of why I've appreciated James' mind for some time now. He has put together a presentation on his thoughts regarding a new kind of Graded Reader that have interested me for some time--one created programmatically based upon the students current vocabulary/grammar progress.
I am no longer teaching Greek, but have long thought that whenever I'm at it again, I will start with inductive reading from the start, but still move through paradigms from the beginning as well. The major shift though, will be that the inductive work through a graded reader is primary, and the paradigm work is secondary. I might vary from James' proposition in that I would not encourage waiting until very long before a critical mass of forms are encountered before abstracting that paradigm, though they can be introduced strategically.
It is very hard for me to still not desire to combine this clause approach with a simplified pseudo-scripture corpus at the beginning--allowing for greater control of when students encounter certain forms/vocabulary.
Don't miss James Tauber's further explanation in the comments section. And if you are ever interested in teaching Greek and still haven't watched the video presentation, remedy that now.