All sounds like its working properly. The 'best target' calculation often will end up selecting the same targets repeatedly as they continuously are calculated to BE the best targets by the same math that led to that last time. Eventually, we could introduce a little randomness to each considered target, but there really isn't that in the AI coding in this portion - and in part that's to speed it up, not that it wouldn't be hardly noticeable.
Of course, that would make it even more annoying because you'd be less capable of predicting where they'll hit next. For at least the feel of gameplay not being a matter of static conclusions for the AI it would be good to add though.
And yeah, it's nice to hear the strategy to save up a bit then come in a harder to manage wave is working.