By the way. I have RX480 and reported hashrate is 21.8 MH/s. According to WhatToMine I should receive 0.0041 ETH/day. Nicehash legacy miner also shows similar but somewhat smaller estimate. But I've tried to mine on suprnova.cc (48h test) and ethermine.org (4,5h test) expecting at least the same if not somewhat higher return, but their estimates are much lower. After the test Suprnova estimated ~0,003 ETH/day and Ethermine currently estimates 0.00274 ETH/day. I dont understand how can it be lower then return from Nicehash when it has higher fees?
The difficulty level will constantly change depending on how many people start or stop mining in that amount of time. There is also a certain amount of luck that comes into play too; luck will vary but should even out over the long haul. There is also the possibility of issues with pools rejecting shares or having downtime as well. Calculators only show what the estimated profit would be if you mined for 24 hours at the exact same settings as it is at that precise moment. It's meant to be a rough guide, no more no less.

I am using Nicehash Legacy miner. Can I copy benchmark values from one version to the newer version to avoid time wasting to rebench all algos?
After the new version of the legacy miner I did a hole new benchmark test.
But some hashrates were lower than with the version before. Does anyone knows a reason for this?
And also, does the benchmark test has an effect on the final mining?
For what is using nhm the hashrates from the benchmark? To choose the best algorithm?
The purpose of the benchmark is to find out what hashrate your hardware is capable of with each algorithm in order for the auto profit switching feature to function. Without benchmarking the hardware that feature can't function since it has no data to base it's calculations off of.