Normalized Dot Product is bugged?

I can't tell if I'm just stupid, or if I'm encoutering a bug.

But I figured these two setups should produce the exact same output:

However, the above two setups DO NOT produce the same output. Setup #1 produces expected results that are in the range of 0.7 to 1. But Setup #2 produces visibly different output, with values that go above the range of 1.