You might be right. I've only had a quick view of the .pdf and accompanying movie. Too busy at the moment to give them a good investigation, but I certainly will do. the idea of secondary illumination comes from a comment from Evans that you get this 'for free'. It's something of an artefact I believe, but it's fairly convincing. Secondary illumination doesn't have to be accurate to look believable.Again, as I understand the Siggraph video, there's no secondary illumination at all, just a very large number of lights. The guy does mention adding a skylight too, but there is nothing that offline CG would define as GI or even as a subset of it.
None of the technologies in the presentation do any secondary illumination, by the way. The first method is a fast approximation of area lights and area shadows; the second, which is in the slides, is a very low frequency fake for ambient occlusion, and the third is a lighting approach that allows for an arbitrary number of lights with a fixed calculation overhead.
Really, LBP's technology could do with a thread to itself (or the existing LBP thread) to investigate what exactly they are and aren't doing. It's worth here is only to say 'LBP isn't easy and Crysis difficult' or 'Crysis is technologically advanced and LBP isn't'. The degrees fo 'technology advancement' various with title and that title's requirements. Evans is nice and up-front about compromises to fit the vision. Every choice is a sacrifice somewhere along the line. For the discussion of 'best graphics', technically it's hard (impossible?) to pin down without knowing exactly what the technical achievements are. The end argument is pretty much pure aesthestics. As would be indicated by three major nominations - LBP's miniworld realism, Crysis' large-world detail, and TB's artistic panache.