I waited a few days before responding to this.

The analysis is the paper is correct, except for the where the POW coin has a tail or permanent block reward sufficient to secure the coin on its own. The notable cases with a tail or permanent block reward are Monero and Dogecoin. In the Monero case the net fees paid to the miners (fees less block increase penalties) could theoretically approach zero under the right circumstances, where there is a sharp decrease in the transaction volume. The obvious example would be on Christmas Day. The point in Monero is that fees are not there to secure the coin, that task is left to the block reward, but rather to regulate the adaptive blocksize and deter spam. I would expect gross fees, before any penalties are paid, in Monero to be about 2 - 4% of the block reward so we are dealing with the case where to quote from the paper In this situation since the fees are insignificant and determined by the block reward and the variance in transaction demand, the financial incentive for a centralizing cartel over fees is simply not there.