Metronome and actual operation times don't agree

We have a very distributed Cloud based infrastructure and we're often called on to measure performance from India based dev teams. We've been basing our performance on populating workspaces and downloading artifacts. Current throughput for data pulled from RTC has recently dramatically improved from ~ 50KB/s to a consistent 750KB/s - 1MB/s based on network improvements implemented by the vendor. However, the metronome results consistently reflect an order of magnitude less throughput. Why would they differ so profoundly?

Metronome ( Average Ping/Upload/Download)     300-900ms | 44 - 250KB/s | 80 - 90 KB/s