Benchmark mode is "average time", so as you can see there was 100 iterations each run,
where I set that each iteration should took 500ms. So when counting that e.g. result = 40 us/op,
during one iteration, there were ((500 * 1000) / 40) benchmark method calls.

You can find more about benchmark modes here:
http://hg.openjdk.java.net/code-tools/jmh/file/27c92e77d4fd/jmh-samples/src/main/java/org/openjdk/jmh/samples/JMHSample_02_BenchmarkModes.java

------------------------
ExistsBenchmark
------------------------
PHREAK:
Drools1387ExistsBenchmark.test  avgt  100  44.277 ± 1.672  us/op
Drools1387ExistsBenchmark.test  avgt  100  42.542 ± 1.207  us/op
Drools1387ExistsBenchmark.test  avgt  100  43.498 ± 1.370  us/op

ReteOO:
Drools1387ExistsBenchmark.test  avgt  100  35.033 ± 1.124  us/op
Drools1387ExistsBenchmark.test  avgt  100  36.377 ± 1.401  us/op
Drools1387ExistsBenchmark.test  avgt  100  35.866 ± 1.131  us/op

------------------------
AccumulateBenchmark
------------------------
PHREAK:
Drools1387AccumulateBenchmark.test  avgt  100  67.007 ± 1.779  us/op
Drools1387AccumulateBenchmark.test  avgt  100  68.365 ± 1.850  us/op
Drools1387AccumulateBenchmark.test  avgt  100  67.269 ± 1.630  us/op

ReteOO:
Drools1387AccumulateBenchmark.test  avgt  100  48.148 ± 0.997  us/op
Drools1387AccumulateBenchmark.test  avgt  100  47.812 ± 0.936  us/op
Drools1387AccumulateBenchmark.test  avgt  100  48.234 ± 0.908  us/op