'I remember a brown car' vs "I remember a brown 1976 Ford coupe with worn tires and a broken side window'.
But would a jury think someone having a 2005 Honda Civic with worn tires and a broken side window is alone reason to convict someone? Probably not? Would they think "A 75% match with [fingerprint equivalent]" is alone enough reason to convict someone? Usually. Because they are very misinformed about what fingerprint "science" can and does do.
But making claims like "fingerprints are good for choosing which of the suspects is more likely to have done it" is harming the situation you describe, not helping. That is, it puts even more onus on the defense attorney to explain how the process is even supposed to work.
If all you have is a finger print--then yea--the possibility someone else with a close match did creates reasonable doubt.
But if the thief was seen in the area at the time and you have a finger print, well that's less reasonable now.
TL;DR; apply Bayesian probability.
Your last example is definitely a good counterexample to this, but I feel in most of cases, the suspects are very few so fingerprints only need to be able to distinguish one of the others, than say, identify a random dude from billions. So it's still useable.
It's like while SHA-1, or even CRC32, is totally incapable for security, it's still perfectly fine to be used to detect if a known file is corrupted (through natural cause, not deliberately altered) or not.