We performed 114% better than PPS pool, at least in last week(even with a 1day 7hours long block).
However it's missing some accuracy since you went from TimeA to TimeB not BlockA to BlockB
There is extra time used to get the first block before 00:00:00 (which makes the total time for those block go up) and extra time wasted after the last block until 23:59:59 (which makes the total time for those blocks go down)
Actually, I believe that measuring between two times is more accurate than two blocks. If you measure between two blocks, you're estimate is biased towards more luck. Look at the case of just one block - in 0 time, we got one block and therefore we have infinite hash power. If you look at any one second period, you'll usually get 0 H/s, sometimes some really large number, but never infinity.
If you look at one block you look at the time from the previous block to the block in question.
i.e. what I said above
Edit: interesting mistake to make that one
Satoshi made it in bitcoin to calculate the difficulty change and no one has ever fixed it
(Edit2: of course it will break bitcoin to fix it like any protocol breaking change ... since the calculation would be different)