Is there hard data on the failure rate of fingerprint evidence? That's quite interesting.
Brandon Mayfield is one example of a high-level failure in such analysis. He was implicated in the Madrid bombings by a finger-print and thus faced serious international terrorism charges despite being completely innocent.
Here's an article that touches on some of the issues, and also mentions some of Brandon Mayfield's experience:
http://www.cbsnews.com/8301-3445_162-4069140.html
A finger-print analyst in the linked article claims 98% efficacy, though there is no data to back up that number. With no standards and no apparent statistics to back up their successes and failures.
For some information see:
National Institute of Standards and Technology, National Institute of Justice, February 2012 report: "Latent Print Examination and Human Factors: Improving the Practice through a Systems Approach" (http://nij.gov/pubs-sum/latent-print-human-factors.htm)
In general there is a move towards probabilistic assessment of the strength of evidence (likelihood ratios), which offers a logically correct way of approaching problems in the forensic evaluation of evidence.
'No studies have proved definitively that fingerprints are unique. Likewise, it is unclear if prints change over time or vary depending on the amount of pressure applied. Statistical models and pattern-recognition software could go a long way toward answering these questions. Additionally, research is needed to expose error rates. The ACE-V method depends on two or more examiners reaching the same subjective conclusion, but doesn't require them to reach it the same way. In one recent experiment, veteran examiners looking twice at the same print came to different conclusions each time. "