Thank you both.

I did figure it would be partly because of the much larger dataset. I had just assumed that something in the execution plan regarding cost (CPU/IO/Operator) would differ more to indicate that one would be better then the other. The hashmatch differs between 66.9272 vs 66.9432 which to me feels insignificant. But i guess there is more to decifering an execution plan then meets the eye.

/T