As others have pointed out, the differences between the results obtained for the two statements is negligible. Even when testing the same statement twice, it is very unlikely you will yield two identical Ops/sec values.
Also noticed you said "a test", not "multiple tests". If the result seemed odd, just repeat the test - don't forget anomalous results could occur if your browser was starved of resource part way through a test run.
JSPerf has always been reliable for me - enough to be useful for demonstrating optimization techniques (this being a classic example).