Your demo is not very legible. Your "earlier" test has a later timestamp, but you have 2 pairs of tests that both happen to start at the same hr:mins:secs ?
This is a test that takes 6 minutes to read from a text? or you have a large number of repeats? When benchmarking things, you must take care to minimize things not being tested. For example, disk access is very slow compared to other things, so to lessen the effect of the disk read, you should have a disk read to put the file into the cache, before doing the timed test. It's really hard to read your test, since the reader is more concerned with how long a test takes, not what the start and stop times are.