Why should a multi transaction of the size of n transactions be cheaper than n transactions alone?
If the amount is the same to each output then it would be much smaller and even if not you only need 1 sig for all the outputs rather than 1 sig per output (so yes it should be a lot cheaper).
Depends which cost function you take into consideration. The size of the block part occupied by a multi-transaction would still be O(n) as are n transaction. So, yes, it could be cheaper by a factor of 2 or so. Calculation of the sig is still O(n), too.