According to Wiki, polygraphs are even said to have a better accuracy rate, at about 61%
I think your estimation of the accuracy rate for fingerprinting is off, 156 people were to identify 7 fingerprints so there were presumably ~1092 identifications potentially made, of which 48 were incorrect. Only 68 of the respondents correctly identified all the samples, meaning that 88 respondents didn't, but since only 48 incorrect identifications were made the remaining ones are presumably instances where the respondent chose not to make an identification.
There are no figures provided in Percy's quote for how many times the respondents who didn't get all 7 right declined to make an identification, but unless it is very high the chances are slim that the accuracy was below 61%.
CTS appear to run tests like this on a regular basis, see
here, and several of these have results where ~90% of the respondents got all of the identifications correct. So perhaps the first year was just an outlier.
TTFN,
WK
™ Version 4.2
