OK. It's far too late. Here goes.
If there was such a way to always weigh 4 balls on each side but do it in such a way that each ball follows its own unique path involving left pan, right pan and unused then depending on which way the scales tip on each weighing we'd be able to work out which was the oddball and its weight.
So, putting left numbers in left pan and right numbers in right pan:
1,2,3,4 against 5,6,7,8
1,5,6,9 against 2,8,10,11
3,5,8,12 against 4,6,9,11
Say the scales went right pan heavy, equal, equal then 7 is heavy.
Equal, left pan heavy, equal then 10 is light.
Try with others to convince yourself.
A property that this system must have is shown in this example:
Weigh 1,2,3, against 4,5,6. Say 1,2,3 is heavier.
Now weigh 4,7,8 against 1,9,10 . Say 1,9,10 is heavier.
As both weighings were unbalanced the oddball must've been present in both. The only ones in both weighings were 1 and 4. From the first one either 1 is heavy or 4 is light. And the same for the second. So there is no way of telling which of 1 or 4 it is. This is only hypothetical but it shows a possibility that must be avoided to make the system work every go. Basically no ball is allowed to "mirror" another's path. So if ball 1 went in the left pan on the 1st weighing, then right then right. No other ball could take the path right, left, left or you wouldn't be able to tell if eg it was ball 1 that was heavy or ball x that was light.
If any ball was not used and it turned out to be the odd one there would be no way of knowing whether it was heavy or light after the 3 weighings. So each ball is used at least once. So equal, equal, equal is impossible.
Also, for reasons that I won't explain but I have scribbled on a piece of paper, no ball may appear in the same pan for all 3 weighings. If it did then another would have to appear in the other pan 3 times and that would be a mirror which isn't allowed. So no pan can be heavy 3 consecutive times.