What you did is flawed on multiple levels. You did two tests consecutively, and then also looked at the payout. The difficulty of the algo you mine will have changed in the meantime, making the 2nd run easier/harder to mine, but never the same. You also looked at payouts but never mention what currency, because if it is anything other than the coin you mine, and it looks like you are talking about BTC, then the market will also have fluctuated the price by the time of the 2nd run. Shortly put: don't trust what you just did ...
It was Zclassic. I infered difficulty from estimated rewards on Zclassic which shows how much of that specific coin you may get based on the difficulty at the time. Yes, I understand this is estimated. Difficulty goes up and down all day so you take a 24 hour average. During the test, estimated rewards differences were less than 5%.
Also, no one has yet to dispute my other point about hash rate when dev fee is on/off. There is no difference between the two option in hash rate shown and actual rewards despite documentation that optimizations are disabled.
Someone else please try disabling devfee.