As I read it, the author performed half the tests with Chrome followed by Safari, and the other half with Safari followed by Chrome, which should have eliminate

As I read it, the author performed half the tests with Chrome followed by Safari, and the other half with Safari followed by Chrome, which should have eliminated any first-order bias. Though I agree that the magnitudes of the numbers are aren't really meaningful, short of doing a full 100%-to-0%-SoC discharge with each browser.