Clearly Expectations have exceeded Reality.
The minimum everyone had the right to ask of BD was to clearly be a better option than Phenom II, either by sheer performance, bang/buck, performance/watt, a combination of the mentioned characteristics or others.
For the desktop user, overall, it isn't.
And the baffling part is that this processor is a monster in die size and transistor count. If the performance was at least in-line with it (when compared at a minimum with Deneb/Thuban) while power consumption and/or cost were high, it could be understandable - the manufacturing process simply wasn't adequate like (and to break the mold of car analogies) the jet fighters developed/produced during the WWII that were inferior to propeller aircrafts mostly due to the manufacturing not be advanced enough to build jet engines.
Or if it was a tiny chip with a low transistor count that had this performance but dirty cheap, that would make sense as well.
Some expectations were surely crazy, although it is always good when human imagination and ingenuity can create something out of spit and wire, but expecting BD to be an overall better buy than Phenom II cannot be deemed as unrealistic.
To the question "why didn't AMD shrunk Deneb/Thuban instead" the answer seems to be "they did" (the "they don't have enough resources argument" can't be used since Llano is a shrink of Deneb+GPU), but Llano doesn't seem to be that good OC (power draw doesn't seem to be that good either when OC) nor does the performance of Deneb/Thuban/Llano seems to scale that great with higher clocks.
The only good things (might be an exaggeration but please read it as a figure of speech) out of BD is that the performance seems to scale better with higher clocks than Phenom II and the architecture is performing so bad in so many ways that it seems to be ample room to improve (this doesn't turn BD in a good buy however).