I agree with the earlier posts. In my opinion, it should not be a percentage, it should be a flat amount of players. Average of top 15-20 for instance. That way, an alliance can't just be a handful of good players, it actually has to have a decent amount, but past that number they're free to take on players who aren't as good yet. There's also going to have to be a mechanism to incentivize not being below the cutoff point, maybe a slight penalty or a cap on your score if you're below the cutoff of members, or exclude alliances with not enough members from the scoring system altogether.
A percentage cutoff in my opinion is better than a flat one, but still flawed. For instance in the current GW2, whichever the percentage cutoff, A-Team would benefit from culling everyone who's doing poorly and leaving only the best of the best, and making a second alliance for the people doing poorly, which doesn't strike me as an intended game mechanic. Even if you don't do that until the end of the game to get on leaderboards, it still strikes me as gamey.
There's also a point to be made about adding a historical component. If alliance score is to be based on airline score, I propose tracking a new metric, per airline: 10 year rolling average of game score of that airline, and then making the top 15-20/etc. in this new metric give the airline score. This does not just reward gaming the system to get as high score as possible at game end, but consistently doing well.