re: those mediocre performance figures
Those were my sentiments exactly after reading the article. The earlier claims made in your reference link appear to have been valid, but as science experiments, not live traffic-bearing facilities. Whenever I detect qualifying attributes in disclosures of record-breaking performance, as "active optical cable", as in:
The V40-850C chip is available in a 250 x 250 µm single die size or as 1×4-channel or 1×12-channel chip array. The target applications are proprietary optical links and active optical cables (AOC) at 40 Gbit/s and 100 Gbit/s.
AOCs employ multiple media elements (in this case plastic optical fiber, but we've seen the same principle at work when using silica fiber, twisted pair copper pairs, and even radio links. In turn, those multiple links are inverse multiplexed to combine multiple lower speed links into a high-speed link. The most obvious form of this is when a new submarine cable boasts of reaching speeds of tens of terabits per second. Of course, here we almost intuitively know by now that the higher speed is merely the sum of speeds on all the individual wavelengths.
Other developments that were being announced at about the same time, primarily in Australia, although I also recall some work being done at MIT and a few other universities, employed a holey-POF that resembled a honeycomb on the fiber's cross-sectional view. Here's an example of one such specimen:
Here we have a case where, in 2008, speeds of multi-gigabits per second were mentioned, but here again e see a case of inverse multiplexing to make it happen. It's not too far a stretch to envision similar principles evolving that use graphene.