1.  Read the article.  Autonomous aerial platforms can refuel autonomously and thus remain airborne for 50-100 hour missions.  Manned platforms of similar dimension cannot do that.  A common architecture is being developed for installation in various autonomous platforms.  Some platforms may be subsonic, others supersonic.  Some may lack A2A capability, others will have it.

2. See above.  Also, if equipped with towed decoys, autonomous platforms can pull maneuvers that manned aircraft are incapable of and this combination may open up new A2A missile defeats.  DEW may add extra capabilities in this respect.

3. You don't pay for pilots, their training, SAR, displays, ejection apparatus, O2 generation, etc.  Savings can be substantial.

4. KAL 007 was a deliberate shootdown, not an accident.  Oh, and what about pilots with itchy trigger fingers?  Recall the eagle drivers who shot down our own damn helicopters over Iraq.  Recall various naval pilots who have launched missiles and dropped bombs on their own ships.  Recall the chopper pilots who shot up our allies.  Etc.

Essentially, any information that can be presented to a pilot inside a cockpit can be processed equally or better by a machine as compared with a human.  External information, such as recognition of visual patterns, is still performed better by a human but not by as much as some might believe and the gap is closing.  (Ever hear of DSMAC?  Primative yet reliable enough that we trusted delivery of our nuclear warheads to it decades ago.) 

I trust I have addressed your points adequately and if not, why, just keep reading about the pace of autonomous platform development ;>)  Time will answer all questions fairly soon.

v^2