Do not use different releases' version of bench to perform comparisons. I believe there is a note to that effect at the bottom of the results figure (unless someone removed it recently.)
If you absolutely must use bench to compare two releases run the older release's version of bench on both. But in this scenario, you're comparing oranges and grapefruit, since you're changing operating systems. They're both citrus fruits, but if you hand me a glass of grapefruit juice instead of orange juice I will notice it at first sip.
