Thanks for your comment @eonwarped.

The distribution of voting could greatly affect the appearance of the curves. If we have the extreme case where every post received the exact same number of reward shares. The reward curves shouldn't matter until some posts started getting more votes. If just a few posts received a large majority of the rewards then the non-linear curves would become quite extreme. N^2 in particular would be a nightmare. The rewards API to be released by Steemit Inc. should be very helpful. I still prefer linear rewards as we don't get these problems or any wild fluctuations.

Edit

Regarding recomputing the curve.

I have not directly recomputed the rewards claim in the formula but I have shifted the graphs arbitrarily based on the value of votes spread across posts valued between 1 and 100 Steem.