DFO uses statistical models to forecast sockeye returns. They give estimates at different probabilities, for example, at the 0.9 probility level, that means that there is a 90% chance that that many fish will return. The 0.5 level will be a much larger number, but there is only a 50% chance that many fish will return. DFO is currently managing fisheries at the 0.5 probability level, which many in the field (including myself) think is not being nearly cautious enough. The predicted 17 million fish is at this level, but at the 75% level (which is most likely more realistic) the number drops to just over 9.5 million fish. This doesn't mean that 8 million fish have gone missing, it's the way fisheries science works. Unfortunately, if forecasts at the .5 level are used, numbers like an 8 million fish commercial harvest get tossed around, and if it turns out there was only 9 million fish, you can say bye-bye to this cycle for a while.
One more thing, to the guys who are blaming the natives for the problems, sorry guys, it's not that simple. There's alot more going on than that.