On the flip side of that, the definition of matched tubes is loose. Some people consider 5% variance matched, while others consider as much as 20% (sometimes more??) as matched.
Then there's the fact that most matched tubes aren't actually matched. Matching transconductance (a common method of matching) isn't nearly as accurate as matching with a curve tracer, and probably nobody selling tubes uses a curve tracer.
I do usually get matched tubes, but the matching I bother with is free (and as I mentioned above, usually meaningless). So whether matched or not, they're probably not. Or maybe they are. Who knows (just kidding, I know, I have a curve tracer).
If you take two random tubes that weren't matched to begin with, and install them in an amp, they might be matched. You'd have to test them to know.
Now what I think is a waste of money is overpriced NOS tubes. If you don't have a really good tube tester, or a curve tracer, then chances are you're getting ripped off (and even if the tubes are good, do they reeeeeeeally sound better?). This isn't an issue of simply matching as much as it is in tube performance, and life expectancy. People love paying for mojo though.
The only way we could really test this theory of matched vs. not matched is to curve trace tubes, and make sure at least 2 are well matched, and a 3rd is not matched within at least 20 or 30%. Then we could swap one of the tubes and do a real A/B comparison.