A year is 24*365=8760 hours. If I put in 100,000,000 NXT and 8760 hours, I get 75,735.24 NXT per year. I have no idea what your figure is, but it's orders of magnitude too small. The return is 0.07573524% pa.
If you have 100 next you would generate 0.11 nxt in a year.
I make it 0.0757352 NXT/year. The return is 0.0757352% pa, exactly the same (to the precision given). Again I'm not sure where your figure comes from; 112 divided by a million would be 0.000112 NXT, so your poor man is getting roughly 1000 times the return of the rich man. I do wonder if your 0.11 should be 0.112 but losing a significant digit.
starting 10.00002% - after 1 year 10.00004% vs starting 0.000010% - after 1 year 0.000010011%
So the poor guy gets a 0.000000011% increase in wealth after one year and the rich guy gets a 0.00002% increase.
I'm not sure what you are doing there. Expressing their wealth as a fraction of the total market cap? Why? My point was, if we consider the ratio between the rich man's wealth and the poor man's, that ratio is not affected by forging. They both make the same percentage return on their capital. If the entire economy consisted of two guys, one with 99% of the coins and the other 1%, after a year of forging the situation would be the same.
The site you linked doesn't take compounding into account, but it doesn't affect this, because they are both at the same rate. It also doesn't take into account the costs of running a forging node, which are fixed and more than a poor person may be able to afford. Leased forging mitigates that. In any case, these are real-world issues of economies of scale that will arise in any currency.