The birth/death adjustment for July 2004 was -91,000 jobs, not seasonally adjusted. Even including _all_ of those jobs in today's number, BLS estimates would be less than half what economists predicted. I hardly think that makes the case for this guy's genius, or against the estimation procedure.
As for the criticism that these are based on "statistical games"--I guess it depends on what you mean by games. The model they use is a non-sampling model, but it has to be, since new firms take a while to be included in the sample. But the model they use is based on historical data and update it quarterly. I don't know what else you're expecting. Perhaps you could elaborate?
Christian