There are multiple well-attested routes around the Shockley-Quiesser limit most of which do not involve just silicon. One, which MacKay touches upon, is multi-junction systems. Another, which he does not (not being an expert in the field) is via so-called multiple-exciton generation systems. Contrary to most engineers' understanding, this phenomenon is pretty well explained from the point of a low-energy continuum QFT for the materials that most care about for implementation. The real game changer would be to mate these systems with topologically nontrivial materials with surface charge transport protection. Unfortunately with current fabrication capabilities, 70% of the conductivity is in the bulk, which does not possess the topological prohibition to one-particle (or one e-h pair) backscattering. (It would also be interesting to see if the known allowability of 2 particle backscattering applies to excitons in these systems)