The geekbench scores are never exactly the same, also the scores on the Geekbench site are uploaded by other users - they aren't "official".....when I test devices I usually kill all apps running, run three geekbench tests and throw them out. Then I run three more and average them.
Does this make my scores more accurate? I don't know. But it seems to limit the variation IMO.
These are my personal test scores (have both the rMini and Air):
-iPad rMini = 1393/2521
-iPad Air = 1478/2682
It would make sense the Air comes in a bit higher because the processor has a higher clock speed - if only slightly.
For comparison's sake, here's some other iPad and iPhone scores I've done. All Geekbench 3.
-iPad mini = 262/497
-iPhone 5 = 691/1195
-iPhone 5C = 705/1218
-iPhone 5S = 1393/2507
And some Android scores.
-Moto G = 335/1073
-Galaxy S4 = 685/1818
-Nexus 5 = 863/2268
----------
This is hilariously wrong....