Bear in mind that the comparison here is against ProRes 4:4:4, which encodes I-frames only. This is basicaclly just compressed whole-frame snapshots--like a slide-show of JPEGs--with no exploitation of inter-frame/temporal similarities.
That choice of comparison codec alone is dubious, as they're comparing an intermediate mastering representation that favours quality over compression to a codec that's meant for end-user delivery. It would have been much more informative if they'd compared h264 against h265, with identical source material and as-similar-as-possible encoding parameters.
