You should run the tests w/out SLOG and with sync=always and compare. From what I saw you ran sync=standard with NO slog and then ran sync=always with a SLOG yet some of your tests in sync=standard may not have actually been using the on-disk ZIL.
Apples to apples would be comparing sync=always without and sync=always with.
If you're doing the test over your network then your network will play a factor in performance too.
Apples to apples would be comparing sync=always without and sync=always with.
If you're doing the test over your network then your network will play a factor in performance too.