If that's the overall expected production for gen3, then an obvious question arises. When is gen4 expected? This year? Next year? Never?
If it's this year, then those numbers seem highly suspect. If the answer is never, then those numbers are meaningless. I'd say gen4 won't be out for a year or two going by those numbers, which would see the chips become pretty weak again compared to the competition. Aren't AM supposed to be working on gen4 already though? In which case, surely it won't take a year or two to be released and therefore AM won't be putting out anywhere near 1600 Ph/s.
Basically, those numbers are completely meaningless without knowing the lifetime of gen3.
No one knows how long gen3 will last, which is one of the reasons for the wide spread of quotes Ph/s. My guess is it will still be a viable (sellable) chip for 9 months to 2.5 years.
As to gen4, that was pretty clearly answered in Jutarul's update:
4) What is the currently anticipated long-term strategy for ASICMiner after the 3rd generation?
re 4) Long-term strategy is to follow the step of Moore's Law, while squeeze all possible efficiency out of each stage of chip design and production.