If a character has been studying fighting arts like Muay Thai, or Klingon Mok'bara for 17 years she is not going to be unable to land a punch in over 40 tries. That's from a real-world perspective wherein Mok'bara would be easily as deadly an art as Krav Maga. And when I did my tests I was not rolling for any particular character - just trying to see how many times I'd fail a roll based on FASA's 96 = Expert skill level.
That's where I get my figures from. In any case I don't think a character should fail that many times in a row with that high a rating. It makes more sense to say the geometric shape of the dice play a part, than to just say player-luck is more important than character attributes and skill levels in a FASA game.
I realize, I'm personally passionate about the Martial Arts thing. And other players might not see it as so important. Others might be more concerned with realism in ship combat, or something along those lines. For me it is when a skilled hand-to-hand fighter looses because of (bad) player-luck that a system starts feeling broken.
Since in my own tests I had consistently better results with the 2d20 than the 2d10, that is why I recommended using them. The only reason I could hypothesize for the differing results in a virtually blind test (no game being played, or character being rolled for) was the different shapes of the dice, and which dice-types were available when the game's system was built.
I might suggest the OP, and anyone interested in doing so perform their own tests - roll 2d20 several times (reading them as 2d10) then roll 2d10 the same number of times and see how many times you roll over 96. Then use which ever dice-type yields results most desirable for your game.
I obviously prefer my Muay Thai practitioners, Capoierists, and Karatekas to win their fights if their skills would normally yield a victory vs the opponent at hand. But don't take my word for it - test it out.