The Block IV missiles can also share targets amongst eachother and decide which target to strike (among a list of programmed targets) themselves based on factors such as proximity and other classified bits.
http://gizmodo.com/this-tomahawk-is-a-mighty-morphin-cruise-...
That said, note that even your example included live updates via comms/satellite links...
Consider the scenario where comms and orbital assets are unavailable/unreliable.
We still need human pilots. We're working to the day when we won't, but, we're not there yet.
(BTW, I'm all for unmanned, and think the F35 is already obsolete, and that humans are the limiting performance factor in esp. fighter aircraft (e.g., limiting it to 10g accelerations, but we're just not there yet for all scenarios, to answer the GP question as to "why not")
UAVs do not yet have the autonomous intelligence to operate on their own, and even if they do, we haven't yet decided to give robots the decision to kill autonomously.
Answer 2 =) we already do, to some degree. E.g., Cruise Missiles have had the ability to navigate on inertial guidance and terrain-following technology for decades, and we use those a lot. But again, a cruise missile can only get programmed for one target, go there, and hit it.
A pilot can go to the target area, notice that the target has moved or new conditions exist (e.g., a bunch of children have been moved in as human shields), or notice that new targets of opportunity have popped up, and immediately take appropriate action.
And, we're back to the autonomous AI tech is not yet that good (and I'm not sure that we want it to get that good).