A case of sloppy benchmarking. Others tried to replicate the results and the only way this could be true was if the original testers had included the loading screen data in the benchmark results.
With the loading screen, SATA being slower takes longer and stays long on the faster FPS loading screen, skewing the results and replicating the original results.
Without the loading screen, the performance is the same within margin of error.
TL&DR: A case of sloppy methodology that included the longer loading screen higher FPS in the results, SATA being slower to load, it was skewed higher.