Just to settle the downward ejection seat topic first, the initial F104As were fitted with downward-firing ejector seats because it was thought that at high speeds the seat would probably be unable to clear the tall fin and tailplane. Hmmmm – nice thinking Bloggs! I wonder where most accidents happen historically? Is it perhaps during take-offs and landings and at low level? Just imagine: “ Bugga – here I am on short finals and the bleedin’ donk’s just lost loadsa power. Oh well – no snags – I’ll just pull up with all my excess energy, roll upside down and eject through the bottom of the aircraft. No worries.” Common sense very soon prevailed and the more conventional form of exiting through the roof was installed. Some 20-odd pilots were killed in the early stages as a direct result of the downward firing seat.
The early J79 engines also caused some rather exciting problems, not least of which was the tendency of the afterburner nozzles to remain stuck in the fully-open (burners lit) position, when the burner was deselected. As I’m sure you can imagine, with the nozzles stuck open and the burners not lit, you would get more thrust out of the back by farting than you were getting from the engine. This caused a distinct tendency for pilots to leave the aircraft to its own devices while they tried to steer the parachute towards a pub car park.
So much for the early aircraft problems. For the overall picture, the basic design led to some interesting flying characteristics. (I apologise to people like John Farley and Lomcevak if my memory of aerodynamics is somewhat rusty, but it has been a while since I had to think too much about that topic very deeply.) The first was the relatively huge length of the aircraft when compared with the tiny 22 foot wingspan, leading to some attention-grabbing, undemanded manoeuvres when playing around with rolling and pulling at the same time (B over A rules OK). This would have been particularly true if you were one of those nutters who believed that the best way of ridge-crossing was to roll upside down as you approached the ridge and pull down over it, then roll the right way up as you entered the next valley. Incidentally, this basically flawed technique also killed several Jaguar pilots.
The next fascinating design feature was the fact that at high angles of attack, the high-set tailplane tended to stall in the downwash from the wing. When this happened the aircraft would almost certainly “swap ends” and would then be quite likely to end up in a distinctly senses-stimulating flat spin. Because of the somewhat exhilarating nature of this possible result of too high an angle of attack, the aircraft had a “stick-shaker/kicker” installed. As you reached an angle of attack that was approaching the possibility of you enjoying a free Alton Towers experience, the stick would start to shake in your hands. If you ignored the shaker, or hit the override paddle on the stick (which disenabled the system) or if you had applied the angle of attack at a very rapid rate, the stick would then be forcibly “kicked” forwards to prevent you causing the aircraft to depart. Now then, the 104 was latterly used in the mud-moving role. So – imagine the scenario: on a range in peacetime sliding gently down the strafe pattern pointing at the target and determined to beat old Fred’s last good score. “Not quite on --- that’s better --- squeeze that trigger --- yesssss --- oh jeez that ground’s a bit close – heave ------“ …… KICK ….. “Bleedin’ ‘ell --- ground’s even closer now!” Choices? Pull harder and get kicker again and hit ground? Hit the override paddle and pull harder and swap ends and then eventually hit the ground after a stirring low level aerobatic display?
As far as losses are concerned, there were nearly 2600 aircraft built in total, and the average percentage loss rate over all the countries that used the Starfighter was probably around 30% to 35%. The early losses could almost certainly be blamed as much on inadequate and insufficient training for the change from the preceding aircraft, and also on low experience levels of maintenance crews and aircrew, as they could be blamed on inherent faults with the aircraft. That sentence is purely my own opinion but – for example – if you buy 916 aircraft (Germany), then service this relatively complex aircraft (when compared with what they replaced) using conscripts, then fly them using inexperienced aircrew then --- as they say --- “you’re bound to lose a few”.
A partial breakdown is: Germany = 32% (although it is worthy of note that their loss rate on the F84/RF84 was rather worse than that); Holland = 31%; Belgium = 37%; Italy = 37%; Canada = 45%; and Japan = 15%. I’m sorry but I couldn’t find an overall loss percentage for the USA. This overall percentage loss statistic is perhaps a bit too broad a figure to be of much use other than to say (“Blooming heck – they lost a lot didn’t they”), as it does not indicate a rate per year of operating the aircraft. I suspect (but haven’t even started to look at proving) that the percentage loss rate per year in the early days of the Gnat, Lightning and Harrier may have been comparable with the equivalent statistics for the 104s. I’m sure that John Farley will be able to either confirm this or raise the “bull****” flag on this statement. Even then, percentage loss rate per year is not nearly as useful as a breakdown of percentage loss rate per flying hour broken down by roles. But that's far too difficult and time-consuming for Boxing Day!
One statistic that I am not even sure is totally correct (but I love to quote it anyway ‘cos it sounds great!) is that it was more dangerous for an American pilot to do an exchange tour with the RAF in the 60s and 70s than it was for him to do 3 whole tours in Vietnam!!
Not sure if any of that helps with the original question, but I enjoyed writing it.