Hey,
Excited as I am about this project, I couldn't help but write down some ideas. Not all of them are original or new; some of them have been mentioned several times here. But If put together, they may have some game-mechanic ideas buried inside them...
On the Samurai unit size
I totally get the idea of Samurai being an elite and rare force and having to trade off the addition of some small elite units against having sufficient numbers to plug the holes in your lines. Everybody playing TW-games likes big armies though; at least I know I do. It seems a little sad to have a 2011 game and reduce some unit numbers back to some 2004 engine capabilities, just to achieve an in-game atmosphere. Increasing the samurai back at least a little and increasing the size of Ashiyaru units to the hardcore cap (?) may be the way to go. Especially if you have research them to recruit them. Also, instead of a raw cap that may irk some players and skew your armies in late-game stages ("I can only recruit as many samurai with 15 provinces as I could with 5??"), a high upkeep cost may be the best idea: the player is free to use them as he pleases, if he has the economy to support them.
On ronin
Even as they are a super addition, they overpower the early game at the moment: it's also strange a Daimyo has 4 ronin units early on, at a time when he is unable to train some of his own clan samurai. Maybe reduce the number of ronin recruitible from early stage sake dens and even keeping katana-units for later stages would be a nice idea, imo. Maybe increasing their upkeep some more might make them harder to rush as well.
On the tech/building tree
It's good to see that you need higher tier buildings and techs to build samurai units. But some units are really high up the tech tree. With the turn-number as they are, it will only be mid-game before I'll be recruiting my samurai archers, and I'm Chosokabe! It's also dangerous to make it so hard for the AI to recruit them: in older TW-modding, this often led to the AI only fielding low-tier units, even in late-game stages... Maybe moving the clan-specific samurai (for instance Chosokabe Bow) to one earlier research tier than the regular ones would be a nice first modification, and put some variation in the gameplay of the different clans! Maybe lowering all first samurai units, but playing with the recruitment time depending on the buiding tier would be a nice approach as well (see further)
On turn-numbers
As always with all vanilla TW-games, I have the feeling the campaign is too short to really use the full potential of diplomacy, inciting rebellions and other stuff the game has to offer. A Vanilla campaign quickly becomes too sheduled ("turn 50: You should be on x provinces by now"). I like to plot more, using bribery to start wars and supporting minor clans to keep my competitors off balance while I build my own power. Vanilla just doesn't give me the time.
Hope you will implement some increase in turns, but I know that leaves much to be changed, from building times and basic economy to tech tree balancing.
On experience/building tiers
I do agree with others experience should come with battle, not the training building. However, we need insentives to build the higher tiers and experience is probably the most important one in Vanilla. Lowering the kill-count for units to gain experience (especially the first few levels) in battle would be cool. That way, you really grow attached to units.
If possible, higher Dojo tiers could be used to increase other stats of the units instead of just giving a raw experience buff, and definitely can be used to reduce recruitement cost/time of the units. That way, your best provinces can really turn out good units at good rates, while less developed provinces take more time and Koku to do the same. It increases the challenge in the early game (long build times for your first samurai units), allows an increase in the overall level of armies as the game progresses and makes your core provinces really important and easier to defend (also for the AI). I would argument an increase in the cost to build higher tier buildings too (I should be an economically important choice to invest in recruiting capability), if I wasn't afraid of the AI no longer building them...