Just tried this mod (my first non-vanilla game in a year or more), and it's a step in the right direction. A Challenging/Hard level start now better lives up to its name, though it's still fairly easy to fend off attackers (more aggressive AI that's incompetent isn't much different than AI that is passive--I guess this is something that must be addressed with core AI rather than modding, which I view much like using bandaids to "mend" a broken bone). Great effort so far, though, with what you have to work with. The unit designs were much more appropriate, and the AI built up their factions quicker than in vanilla, which gives the false impression of a challenging game to come. The end game is still a letdown as the AI squanders all its resources and falls apart under even a casual assault.
Unlike what some misguided people claim, it has nothing to do with AI's inability to handle magic--it really is an optimization problem in other areas (areas that the GalCiv2 AI handles beautifully). The only magic I used in this test game were city buffs and Arcane Monolith (I dislike using pioneers for outposts); the rest was simple exploration, exploitation, pushing research, and then ramming it down the throat of any near AI that declared war on me just because their "power" score was higher. If the AI did likewise, they be far more successful without having to figure out how to use complex magic setups.
On the off-chance that you modders have powerful enough tools to tweak behavior, I'll list what I noticed:
AI still did not optimize units for initiative, fielding mostly 15-18s against my 26-ish units; they should at least have a few units optimized as first-strikers/vanguards. They did not coordinate attacks on anything of value (collectively, the 3 factions I wiped out captured 1 of my outposts for a couple of turns), or try any threatening tactics.
Then when I counterattacked, they did not protect their cities, even with units that were close by when I entered their territory, so wiping out their manufacturing power without any losses was without satisfaction. Even losing factions need to put up a real fight for the victory to be worth anything; pull back units to any city that has enemies close, pump out troops instead of non-relevant buildings (they are getting beat and they are building a garden?). Truthfully, a good boost to their AI would be the distinction between peace and wartime; they should be optimized to build size and economic strength before they ever declare war, they should not declare war before they have a good handle on their infrastructure, and then they should go into full war-production mode once in a struggle with an opponent that's even close to their power rating.
As many people have noted, they still made little use of the power they built, squandering what lead I had given them by inefficient or seemingly random pathing (the AI should be taught that when they declare war on the player, they better be really serious about it, or they be made a laughing stock).
A couple of possible bugs I noticed that aren't in my vanilla games: Sov/champions get dodged fairly often even when the creature has a 0 dodge and the attacker has over 100 to-hit (in one case I noted, my Sov got dodged with a 164 accuracy vs. a 0 dodge--this should never happen). If this is done intentionally, then it's a perfect example of artificial "challenge" that gamers hate, because it's cheap/cheaty/unsupportable, rather than being something that presents a genuine challenge.
All of the factions "met" me early in the game, even though they were far away and I never saw a unit or territory from them (I frequently use the cloth map to easily spot where they come in a vanilla game from so I can plan my expansion/defense accordingly, and in your game, they were nowhere I could see). Insane Ceresa even declared war on me fairly early on and then never showed up--she was at least 30 movement turns from my closest territory/unit (it took me half the game to find her). This is one of those bugs-within-a-bug problem (allowing bad diplomacy programming earlier in the game is only hurtful). It's far better from a game tactics standpoint to not have "diplomatic" relations with a faction you have not actually met. Unless there's something I'm missing about this, early contact benefits the AI none at all, and therefore provides no added challenge for the player (in fact, I used the early contact to more easily control the opponents and my strategy).
A couple bugs I have not encountered in the last few versions of vanilla also popped up: the build queue inaccuracy and the inaccurate move distance calculation (though I have no idea how your mod could "bring" those old bugs back).
I'm still hoping this game can produce a challenging end game, to balance out the fun exploring, amusing quests, and often tough starts (I keep terrible starting positions for the extra challenge of coming from way behind). All the great tactical games demand sacrifices of the player (you will have to lose units to achieve overall victory rather than the "never lose a man" victory that FE has). If you want any more feedback, let me know; I consider myself an average 4x gamer, so should be good to test ideas against. In any case, I appreciate your efforts. Happy modding.
ETA: To clarify, I usually play on a Challenging world setting, with a mix of Hard/Expert AI (all Expert gets annoying since they have so many boosts to money and research). Do I need to modify these settings to get a better feel for your mod?