As Krom said tower accuracy has been increased quite a bit since the last build. The difficulty with balancing it is that most of the inaccuracy doesn't come from the random element we add, it comes from the fact that units change direction. If you change direction while the stone is in the air, it will almost certainly miss you because it's aiming to land where you would have been had you continued in a straight line. So we can't look at our code and say "on average 3 out of 5 stone will hit", instead we can say something like "if a unit is walking in a straight line, 4 out of 5 stones will hit" but that's pretty meaningless because units don't always walk in straight lines
Also please keep in mind that towers will miss fast moving units more often than slow moving ones, and slow moving units more often than stationary units. This provides an incentive to charge at towers with axefighters/militia (rather than walking) or to run cavalry at towers. It adds more tactics and variety IMO.
The best solution I can come up with is we try to get it "feeling" right to us in tests (which I've already done) and then during beta testing of the next release YOU guys tell us what you think of it, and from that we'll judge the community's view on the matter. We've done this in the past with stuff like archers accuracy, which most people seem to like now (except that some campaign missions are harder to win, which we're fixing separately)
So, when the announcement of beta testing comes out, contact us if you're not already on the mailing list and test it out for yourself so you can tell us what you think. It's just not possible for Krom and I to test balance issues thoroughly, you need to play real matches with real people and see how it goes.