I can't think of any sensible way to decide whether or not data was "generated by" a piece of software. If you export your book to plain text, is that not "generated by" the app? Why not? It's the app's output. But it's also your own copyrighted work in a more or less rudimentary form.
Imagine you first write your book in Apple's app and save it to an iBook file that falls under Apple's restrictions. Then you retype your entire book in a 3rd party app that can export iBooks perfectly. The 3rd party generated iBook is bit-for-bit identical to Apple's. Now you have "two" files that are subject to different licenses, except of course they are the same file.
That's already absurd. But what if instead of retyping your book, you just import it into the 3rd party app from the Apple generated iBook file. You should be able to do that, right? Surely the legal status of the data can't depend on whether or not you typed it by hand. But again, this is absurd because you are importing and exporting to the same format. It optimizes to a file copy.
And this isn't a reductionist argument. The above is an entirely realistic scenario that will inevitably come up whenever an iBook author wants to sell their book outside of Apple's store.