I doubt this is still relevant, but I'll comment regardless since we've seen similar issues. Obviously, the New Relic transaction overview is not drilling deep enough, since the filter() method encompasses the entire request processing.
If you were to look deeper, you'd probably find that a) you're either doing something wrong in your code (eg. inefficient SQL, non-parallel calls to resources, ...) or, b) you're encountering a lot of blocked threads that are timing out (hence the value so close to 60 seconds - that looks like a timeout) on some OGNL methods or resource bundle handling. I suggest adding a profiler and having a look at the blocking of threads.
If you're on 2.1.8, you might want to a) upgrade to a later version, or b) apply the patch that was provided to resolve the several blocking threads in the OGNL runtime. We've had similar issues, and applying the OGNL patches and backporting the locking of the resource bundles improvements made a massive difference in the processing time of our calls.