40% of players completing the game isn't that bad. First, you have to understand the nature of Steam - a lot of players never play a lot of games thay bought. You know, discount sales, much cheaper, I'll play it later at some moment, etc. I know the lad who has about a hundred purchased games on Steam he has never played. Most of my friends have some game they ve never played. I do.
Secondly, there is a low chance to retain a player who won several Settler games only, meaning make him buying Civ VII or another Civ VI add-ons. Almost all games even the most successful ones have qute low players retention rate. Most players buy (or download for free), try and quit. Or try for a little bit longer and ... quit. I would love to see how many players have won 10 games, 50 games.
What I mean is that developing a great AI for guys who love playing diety is a long term investment - features like this guarantee that these core players will stay buying all DLCs, add-ons and subsequent games for years.
What a smart game development company does is logging everything all their players do, analyse it and make right decisions based on that. It's not just only 40% of players have completed a game - it's literally everyhting you can imaging. So the 40% fact by itself shouldn't affect the decision to improve the AI. On the other hand, even top companies fail doing their job from time to time and we can only guess what happens in Firaxis.