The house in Vilonia that had both its exterior and interior walls bolted down, and the well constructed homes in Chapman (one of which was swept away with such force that a part of its poured concrete foundation was snapped, and another that basically disappeared with no debris remaining) seem like pretty reliable EF5 DI's to me.
Even taking the small sample size into consideration, the point of that study was still this: from a 'pure' engineering perspective, there were very few to no structures along the path capable of indicating an EF5 rating. Obviously, any tornado that causes genuine EF5 damage is going to do it in a small area relative to the rest of the path, which makes sense. Still, the ASCE's logic is not a good way to approach tornado damage surveying, especially since most structures in the US are poorly built and surveying taking only engineering into consideration sets an impossibly high bar for EF5 ratings.