Another example - it would be a waste of time for me to compose 10 minutes of highly interactive music based on the speculation that a client would be able to implement such a system. It would be better to do 1 mock up to show the intent and direction I'd have, then wait to ensure that such tech actually came to light and worked. I once worked on a project where they wanted 7-8 tiers of music for interactive music via the iPhone. This was several years ago so some of the tech hadn't really been fleshed out to the point it is now. So I wrote the music and broke out all of the tiers.
In the ends, the client couldn't make it work and I ended up wasting some time. (Not a lot but some.) Thankfully, it was just ONE song. Imagine if I had done all of that work upfront for a whole sound track, only to find out the tech got in our way. This is why I say it's better to let most of the game be set in stone. So you know exactly what you're dealing with and what's possible. But, again, I agree about setting direction and intent early on.