Testing 101: Change one variable, retest. Observe differences.
You have changed one variable, network. You have observed a significant difference. My working hypothesis is that your change in throughput is directly related to the one variable changed, the differences in the available bandwidth between generator and target from location a to location b.
So, your application proves to be more scalable than your network infrastructure in this case. It also goes to point out the challenges of testing across a network of unknown quality and scalability to test an application of unknown quality and scalability. Keep one of the factors as a "known good" when trying to assess the qualities of the second item.