I just did a simple test of one belt between containers, for 1 minute per belt, and the follow numbers result:
Mk1 (60) belt: 60 in 1 minute
Mk2 (120) belt: 120 in 1 minute
Mk3 (270) belt: 270 in 1 minute
Mk4 (450) belt: 450 in 1 minute
So the belt numbers are spot on, I think the splitters are doing a ratio of 1/2/4 on the 60, 120, 270 belts, effectively making the output of 60/120/240 not 270, and the remaining 30 would be added (4 or 5)/~9/~18 to the belts but take longer than 1 minute
Just tried the 60/120/270 split and the splitter is definitely not working optimally with those belts, if you want a cleaner output you'd want to split the 450 into 2x 225 (on 270 belts) or 3x 150 (on 270 belts) and go from there.
I don't believe splitters will work optimally for uneven splits (i.e. splitting 60/120/270) which means perfectly splitting a 450 belt to 30 belts if not possible as you can only perfectly split by 2 or 3, and since 450 is 2 x 3 x 3 x 5 x 5 and you can't split by 5 perfectly with the current tools it's impossible. Bear in mind that with the oversupplied machines backing up the system will self balance so you shouldn't lose too much production as a result.
I would appreciate if the devs put in a special case on the splitter logic for connecting 450 in 60/120/270 out since it looks so neat to do that and would allow us to split by 5 implicitly (2/4/9 ratio) and make the numbers nice and neat, but I imagine it's a low priority.
I found out the splitter doesn't care about the speed of the belt when it splits if you put 150 items in, and have a 60 belt and 2x 120 belts out, all 3 belts will get 50 items, so the logic must try each output in sequence and skip an output when that output is still full, which would result in suboptimal splits when trying to not even split by 2 or 3.