I've noticed this too. Were the numbers not being jiggered, then they should - more or less - be revised upwards 1/2 the time, and downwards 1/2 the time. The fact that they're always revised upwards is evidence that they're being played with, IMHO.
It just means that the economists' models are wrong.
(You realize that the "unexpected" comes from the polling of economists to determine an estimate...it's not the government giving it as "unexpected," right?
So what we have is a statistically significant indication that the economists' models are ridiculous.