I already saw that and made the connection to Windows 8, but drew the opposite conclusion.
First, the beginners shell vaguely resembling the start screen shows that a larger presentation like that was seen to have value even before it was motivated by touch. Second, some of the decisions around the start screen seem designed to address those points - it's a complete replacement for the start menu, not an additional "beginner" version from which you're eventually meant to graduate, and there's a class of apps that do look and act like the start screen.
Another relevant blast from the past here would be this Windows 7 blog post on touch:
"While providing this touchable experience, we made sure you are getting the full Windows 7 experience and not a sub-set just for touch. We've been asked if we are creating a new Touch UI, or "Touch Shell" for Windows – something like Media Center that completely replaces the UI of Windows with a version that is optimized for touch. As you can see from the beta, we are focused on bringing touch through the Windows experience and delivering optimized touch interface where appropriate. A touch shell for launching only touch-specific applications would not meet customers' needs – there would be too much switching between "touch" mode and Windows applications. Instead, we focused our efforts on augmenting the overall experience so that Windows works great with touch."
My take: Yes, having two kinds of applications is a UX penalty. However, the penalty isn't infinite and can be justified if benefits exceed it. In this case I think having the new UI / app model be the "beginners shell" AND the "touch shell" AND the "instant on environment" (around 2009 when Windows 8 was being planned there was a trend for netbooks to initially boot into a Linux environment for basic internet browsing and status checking etc., before going into a heavier Windows environment) AND help introduce a new application model, together those benefits are enough to justify the cost of having a second model (and of course the long-term goal is to make the other model unnecessary for most people).