Three systems treated the same, have different profiles. PG profile was low enough, the benching had to be altered, to get the results that they were expecting

Three systems treated the same, have different profiles. PG profile was low enough, the benching had to be altered, to get the results that they were expecting (or the reader to be expecting). If it was a technical implementation issue, why wasn't it done right the first time? Probably hubris, in an obvious case of tweak the testing methodology until you get the result you wanted.