Therefore, they build starports everywhere to make use of the manufacturing points, which would otherwise be unused and returned to their treasuries (which the AI percieves as bad).
Nice insight.

Your guess is probably as good as anyone else this side of Brad's.
But all in all I think GW's observation that Starports on most every planet is a necessity in a short game and the AI's merely do the same thing regardless of game size.
Could it be that a Starport everywhere makes sense from a short-game POV and the AIs can't change behavior when they're on Huge and Gigantic maps?
I used to play gigantic abundant all games exclusively, but I've recently come to appreciate some of the idiosyncracies of the smaller game by playing Metaverse League games. It's not always obvious how a strategy translates based on size but certainly the concept of planet specialization that works so well when you have a large number of planets does not translate well at all to smaller maps. This actually came as a shock to me that required quite a bit of scrambling to recover from.
I would guess that if you're going to have one basic strategy that's used for all game types the safest one to implement would be the general approach where each planet has a bit of everything; production, research and income. As millertime points out production equals ships and ships require starports.