The biggest problem I have WRT estimation is how much I can do things my way, not necessarily the quantity of features. If I can do things may way I can crank out stuff pretty damned fast; usually faster than management can obsorb whatever new questions or decision points that may come up along the way.

It is when people put clamps on the way I do things that things slow down. For example, dictate their favorite architecture (often out of habit, not nec. merit), dictate "odd" schema designs, etc. An example may be an ex-mainframer who wants to do things in a mainframe-like way.

That does not necessarily mean that I won't do it their way, for if they call the shots, they call the shots. But the end result is that it slows me down.

Features are secondary to estimations.