No, I meant what I wrote. The time spent in total in GC includes the time spent performing allocations. If you reduce the time spent in allocations with all else being equal, you reduce both latency and increase throughput. In that way, optimizing for your allocation latency also increases your total throughput.