Tangentially related anecdote: 15 years ago, a friend of mine bought a house on the (wrong side of an) edge of a crime reporting district in Philadelphia for $80k. Across the road (which wasn't a highway or anything, just another inner neighborhood road), an exact copy of the same house (same developer, same year, same project, same layout, same everything) went for $200k at the time.
Three years later, the crime reporting districts changed, and his house (and a few others) were reported in te better district. His house value changed, practically overnight, from $100k to $240k.
Ergo, report methodology can effectively affect related value significantly. Quantum measurement effects on a macroeconomic scale.
BUT this article was very unsatisfying. The above factor is just one of many. And the problem is that paramilitary "boys' toys" enthusiasts, prison guard unions, recipients of anti-terrorism grants, and all the other interest groups are all contending to lead the low-crime parade.
It's pretty certain that some crime is becoming less prevalent. It's hard to hide a shooting or stabbing victim. But these gains have to be defended against the kinds of fudging you describe. And, to reap the social and budget benefits of real reductions, we also have to call BS on false claims of the lock-'em-up interest groups.
It could be that the answer is that more people know what a normal middle class life is like and therefore more people want to live that way, even if it is just aspirational.
And this got me wondering - did crime rate indeed go down? Or is it just that numbers are measured differently and thus look better?