To achieve the results with NH, your gear is obviously being rented and the average rental price is higher than expected BTC earnings. I'm curious to know what backup pool you've set on that miner, and if you've configured it to have a minimum accepted payout rate on NH (by setting your password to something like p=0.011).
By the way... you've certainly proven that your first assertion of "no such thing as 100% payouts" from your first week observations is incorrect

.
I'm not using any p= parameter for nicehash, I'm using "x" for password.
And as far as backup pools - I'm using backup pools for the same pool if it is available (for example, usa and eu stratum servers for the same pool).
If pool would go completely down (all it's stratum servers goes down), then an additional backup pool is set to nicehash stratum since it pays on PPS and it looks like best paying PPS so far -> but this is set on a different payout address (not the one used for experiment). This would also mean that if one pool would go completely down, miners would still be hashing away (I don't want to waste them

). However I'm also monitoring miners with miner's dashboard (
https://github.com/selaux/miner-dashboard) and I'm barely seeing any switches to this last-resort pool, therefore I can say that all pools, chosen for this experiment are very very stable, near to 100% stable.
And yes, I'm happy to see that 100% payouts can be achieved, we'll see how will this go on long term, I'll do another comparison to theoretical payout after the next diff adjustment.