They announced in June but Intel had already announced that Thuderbolt 2 was not ready ahead of late 2013.
Image
[ from a story on June 9
http://www.anandtech.com/show/7049/intel-thunderbolt-2-everything-you-need-to-know .... just a couple days before Apple's dog and pony show ]
Intel's plan for more than a year was to ramp TB v2 in 2014, not 2013. They pulled it bit a bit, but once the Mac Pro was coupled to that, large scale production couldn't start until very late 2013.
E5 v2 supply availability was too much better than TB v2's through most of Q3-Q4.
The announcement in June was far more so for folks to buy the old model if they are deeply attached to "box with slots" design. There was enough even with extended purchase order submission constraints to get orders in.
The announcement were of a system where the parts weren't even available yet.
There are no indications that the shortage is not simply demand exceeding the designed capacity for the factory. If Apple contracts for a 5-7K/month factory then when 14K folks show up on in first week it is going to take time to "dig" out of that hole (even if started production a month ahead of time). Even longer if another 8K show up in the next 4 weeks after that.
Over the course of the likely 10-15 month lifetime of this specific Mac Pro that is likely closer to the correct move because the wave after the initial bubble probably isn't going to be that high. If Apple goes on to sell over 85K Mac Pros this year than yes this was a blunder. If it comes out to around 72K then it was right on target.
The other fundamental problem is how do you build thousands of CTO machines months before people order them. Ouija boards and gypsy fortune tellers ???? With an substantially mature product with an well established ordering record they could take a very educated guess. For a relatively brand new product, it is basically a crap shoot.