A test should normally run 5 iterations of warmup, followed by at least 15 benchmark measurements.
Above, the "at least" part means that a ScalaMeter Measurer can decide to run the benchmark snippet more often.
This measurer exists to provide more stable results, so that a performance regression gets reported only if there is a real regression, and not just random jitter.
My guess is that, what might be going on, is that the OutlierElimination measurer is repeating some measurement.
If you turn the verbose -> true configuration option on, you might be able to see this.
In this case, the solution is to override the measurer, like here:
Also, if you provide a runnable example, it would be easier to give you an exact answer.