An example of consistent results might be a case where you use the UNIX system timer, dtime, to measure the work performance. Suppose you have 2 different tasks whose work is measured by the number of operations (OpCounts) for a problem of size N. One task is expected to have work O(N) and the other task is O(1) or constant. If you obtained computed results indicating that as N -> 2N the work(2N) is roughly 2 times the work(N) for the O(N) task and roughly unchanged for the O(1) algorithm, then you have a consistent result and this helps convince the reader that your data is to be trusted. But both OpCounts need to be supported, not just one, if this is to be a consistent result.