I think it’s possible for both to be true.
Most beastmen would be in villages, because they don’t have a leader capable of uniting them into a semi-civilization (I say semi because they also have weak farming abilities meaning lower income, combined with the maximum population limits per hex, would mean that a beastman King would need to control much more land area than a human King to reach the same level).
But the big world-destroying armies come out of those rare semi-civilization areas.
Composition is still a bit of an issue, I agree. Have you considered using the standard mercenary rates and converting them to analagous beastmen?
For example, goblins might replace light infantry, goblin wolf riders replacing cavalry, hobgoblins replacing heavy infantry, ogres replacing cataphract cavalry, etc. I don’t think you would get a perfect 1:1 correspondence, but you could use them as a starting point.
Alternately, you could just say it’s a demographic thing and start with the demographics of the area. If the domain is 80% goblin and 20% hobgoblin, you could say that there might be an ogre here and there, but not enough to make any meaningful units. (Although they might have an ogre lieutenant for some units.) Then you could use the D@W breakdowns for the goblins and hobs separately, simply based off the number of each that happened to be within the domain.
I think in general, ACKS assumes that a single type of beastman is dominant over the others in any given area, the same way that a realm is a human or elven domain instead of being a mixed domain. Which makes sense; most types of beastmen are happy to war against or enslave other types of beastmen, they’ll only stop fighting each other to fight humans when the humans are currently in the process of invading them.