Xavier is dead on about that. Endpoint DOES use the Boost regex engine, whereas detection servers use the Java engine. In MOST cases, these are pretty similar and that regextester.com site gives a pretty good representation of the results you'll see with both.
I recall running into problems with some differences between the two implementations of regex with a customer of mine some time ago. It had to do with the support of positive look aheads (or maybe it was negative look behinds) with Boost (one or the other didn't seem to be supported), so be on the lookout for that if you're using those structures in your regex. To date I have never found a good online tester for the Boost implementation of Regular Expressions.
Hint, and what I've been playing with lately...you might be able to accomplish what you want better with a Custom Data Identifier with a Custom Script Validator (available in V11 now). The scripting language (basically a very limited implemenation of Perl, and descibed in the Custom Detection Guide) will allow for additional validation of the match as built into the Custom DI, and would work consistently among Endpoint and Detection Servers. It may even be more efficient than a Regex.
~Keith