Thanks for your response @miniature-tiger. Figures 1 and 5 are looking at slightly different things. Figure 1 shows the point that upvotes on a post are worth the same as the linear rewards curve. Whereas, Figure 5 shows the point where total payout of a post catches up with the total payout of a post upvoted using a linear rewards. As the early upvotes are worth less with the 2e12 curve, it takes sometime for the upvotes to lift the value of the post to what it would be if we used linear rewards curve. Based on the assumptions I have used, this might not be until the post reaches 54 Steem. It is possible this could be sooner or later depending on how people vote. The shapes of the curves change based on voting behaviour that's what makes them difficult to map using Steem. I guess that is probably why Steemit Inc. discuss the formula in terms of reward shares. Unfortunately, the majority of the users do not understand reward shares and an explanation of the effects in Steem is considerably more useful even if the exact numbers are impossible to predict.

I agree with your two conclusions. Your second conclusion is probably more concerning. Some users may choose to bid bot there posts to extremely high values to make as much as possible. It will be interesting to see how the bots respond to the new curve as well as all the other changes. I think it will be very difficult to calculate an accurate ROI as it will vary from post to post.