The big reason we don't do F2F in our data collectors is that they are too old, and do not have that capability. Eventually, we'll upgrade, but that isn't in the cards for the immediate future.
As far as creating everything in the data collector, I've heard varying feedback on that route. We haven't tried it yet ourselves, though, so I can't give a first-hand account. I haven't really noticed much benefit of having "smart" 3D objects from field data import - actually, that creates problems. So we don't care so much about that, except for the aspect of adding breaklines to surfaces easily (something that is easy in C3D, as long as you only work with one surface per Survey Database).
The big complaint I've heard of the F2F solutions is that, when the guy in the field is trying to create what is essentially your final linework, it takes more time. I can't corroborate that from first-hand experience, though. But even if it takes only an extra half-hour for the field guy to complete an 8 hour survey, then it isn't worth it for us.
Say we send a field guy out and he spends 8 hours doing a topo survey. With C3D and its linework coding, when everything is working and the field guy knows what he is doing, we typically have complete linework inside of C3D, along with a fully-built surface, in 15-30 minutes. When everything is EXACTLY right, we have it within 5 minutes.
We've heard from multiple sources that, when the guy in the field tries to create these complete drawings, it ends up slowing the field work by a good chunk. And if it takes a field guy 9 hours to do that Survey complete with linework drawn in the data collector, it already doesn't make sense, because with C3D and fieldbook coding we can do the same thing in 8 hours of field time + 30 minutes max of office time. Actually, even if the field guy takes EXACTLY the same amount of time, and can complete that survey in 8 hours complete with linework with no office time, it is still a borderline call. We would have to buy new data collectors and software, and train the field guy to use the new stuff... All to cut out that 5 to 30 minutes of office time per job. That's a tight call... Is that really enough savings to justify the expense?
I should stress again that we haven't actually tried one of these F2F solutions. I'm open to arguments, though, because we'll be hitting the point soon where we'll want to upgrade, and we're interested in the solution that works best overall, in real-world usage in real survey firms. We are not locked in to any one thing. But our current solution leverages the strengths of both our office and field personnel, and works quite well.
It could also be one of those things where the ideal solution varies. I've heard that some one-man shops really like the complete solutions that do F2F in the data collectors, because it's the same guy doing all the work in either case. Don't really know myself...