Something else was running in the background during your disabled test because it makes zero sense that using a feature that creates more reads and writes to the nand flash would actually increase a nand flash read and write benchmark performance. Logic dictates that your test is a fluke/invalid.