It's not simply a hyphenated word. It looks like two long sets of characte= rs with a hyphen in the middle, which is the exact same thing as a unique id.
The rule doesn't do very well anyway:
1.039 1.1433 0.1190 0.906 0.73 0.90 SUBJ_HAS_UNIQ_ID
Hence the <1 score it receives.
Perhaps this is a decent chunk of why the rule doesn't perform well.... It might be worth looking into modifying that regex in the eval to try to get better performance, or splitting them up so you can test each separately...
