Well perform a test, change maxmem to 1.5GB & re-run, afterwards set 2G maxmem & re-run.
For me this is common sense something to grasp something is not right. Either the Top 5 in that ranking suck monkey balls or you are too darn good...
I fail to see the optimisation of the benchmark when you boost 80% in memory score and you beat way higher clocked setups. It's intentional bugging the output...
one does not jump from 11K to 18K in memory score without bugging the benchmark... Provide the maxmem setting plz ?
I hope you don't believe that your parameters allow you to beat scores with the same hardware clocked 500-700Mhz higher...