I agree that getting finger input right was the critical factor, but I don't think it was at all obvious that it would produce a completely different user experience. Looking forward from five years ago, I don't think you could call it anything but a tweak---it's only with the benefit of hindsight (unless you share Jobs' unique talent) that you can see that a small design change (use your finger instead of a stylus!) could make such a big difference.
Take something simple like zooming or scrolling a page. In Windows Mobile circa 2006, the basic paradigm was the same one you'd use on the Xerox Alto in 1973: click on scroll up/down buttons to scroll, click on a menu and then on zoom in/out to zoom. The stylus was just a less capable mouse on a smaller screen. On iOS circa 2007, you pinch to zoom the page and swipe to scroll it. There is no analogue to this in the previous Windows/Menus/Icons/Pointers paradigm.
I think the iOS UI is really the perfect example of something that wasn't just tweaking. Windows Mobile was just a tweak of the mouse-based Windows UI for stylus input. iOS threw away the Mac UI, predicated on the mouse, and started from scratch. "What does a UI look like when you start from the assumption that you're manipulating it with your fingers?"
Your contention earlier was that stylus -> touch was a "tweak." My point is that it's not just a tweak, it's throwing out the existing model and starting from scratch. You implicitly acknowledge that by referencing the investment to make the change, an investment that is incurred by having to throw out the existing design and start over.
Touch was not a tweak. It was throwing out the steering wheel and asking "so now how do you drive the car?"
Exactly my point. Let's say it drives by telepathy. The car wouldn't change much: it would still have a driver and be driven on the road. It would still have an engine, passenger seating, storage space and cupholders. It would still at a glance look like any other car. But that apparently small change would result in a profound change in the user experience (and yes, profound engineering challenges for the maker). It's a better car. But it's not an entirely new class of transportation, just a tweak to an existing one, like the automatic transmission and hybrid engines.
I'd argue that the Zen Jukebox did not prove that HDD MP3 players could be successful, and that Treos and other smartphones, though proven successful, were not substantially in the same form as an iPhone, and that previous tablets were proven failures and also not substantially in the same form as iPad's.
I think the iPad is really the best example of Jobs' talent being used in a capacity that wasn't just tweaking. Stylus input and finger input are just completely different ways to approach the user interface. An iPad isn't a tweak of previous failed Windows tablets, it's throwing out the basic premise of the machine ("a tablet is just like your computer but in the palm of your hand") and starting from scratch ("what is the purpose of a tablet?").