In my opinion the difference in speed is barely visible, you seem to have a very flat optimum here. Since you did several test runs with different dev ratios, can you tell us from your data at which ratios you saw differences which would be more than let's say 1/3 stop or 0.1 density?