From a technical standpoint - your friend's iPhone 4 could have any IP address at all. There are only two ways for your phone to find out that address in order to initiate a FaceTime session. 1) Contact a central server that ever iPhone 4 contacts in order to tell it its IP address (this is how most IM programs work - you log in to a server.) Every indication is that this is NOT how Apple programmed FaceTime to work. 2) Since the phone company DOES know how to contact every phone on its network, your iPhone will have to make use of the cell network to establish an initial communication to exchange IP addresses, with which the FaceTime session is established.
So at some point, it seems to me 99% likely that a cell network connection IS required, there's no other way for your phone to know how to contact your friend's phone - it's immaterial whether you initiate a normal call first then hit the FaceTime button, or just hit the FaceTime button first, a cell connection is likely established either way. The question is, does it count as a voice call? If so, how long? A single minute, or for the entire conversation? Or does it count as a background type notification like how we get notified of a visual voicemail? Or does it count as a text-message?
Not like it really matters to speculate about it now - it is what it is, and we'll get more info about the mechanics when the phones are actually in users' hands. 🙂