The comparison is not completely useless, because it is possible to estimate the hashrate the supercomputers could sustain and compare that to the current bitcoin network. Then it is comparing apples to apples.  The comparison done is going the other way, trying to estimate the FLOPS rat of the bitcoin network's hashrate.

I think the comparison is interesting in that it gives us a good idea of how much impact th eexisting supercomputers could affect the bitcoin network if they were reprogrammed to mine bitcoin efficiently.  The fact that even with all 500 top super computers turned on bitcoin, they could capture about 10% of the hashrate is rather interesting and somewhat comforting that we are beyond the possibility of a 51% attact by these supercomputers. If somebody was to try a 51% attack, they would have to build the system from scratch and not just repurpose an existing system.