Disagreement in performance results

So we have two separate connections. An ethernet ATT 100/100 and a lab network provided by the enterprise which is part of a MPLS circuit that goes from our site to STL and then to charlotte to hit the internet. It’s 300Mbps to STL and supposedly 1Gbps to Charlotte.

We are having major discrepancies in performance and aren’t getting good answers from the enterprise. Anything we download from major provides on our ATT connection we get a static 100Mbps/100Mbps up down. So for example I’ll do a iperf to iperf.he.net and will get like 97Mbps.

Doing the same tests on the enterprise connection we will get anywhere between 2Mbps and 15Mbps. It will bounce all over the place.

I’ve tested multiple sources (softlayer, att, dslreports, google, etc). Every single one I’ll get around our 100Mbps on our att connection and then again on the enterprise connection it will be all over the place and jump from a couple Mbps to 20Mbps. Usually on average I’ll get 3-4Mbps.

I’ve told them that it’s basically unusable for us. They keep saying nothing is wrong and quote that our TCP windows are too small or they run a UDP test between their sites, get line rate and say nothing is wrong.

I’ve validated that the TCP windows aren’t the problem as I’ve checked it and it will be 1-2MB in size for the TCP window size and a RTT of 50-80ms. This would give a theoretical max of 150Mbps...which shouldn’t be limiting it to 3Mbps avg. As for the UDP testing i’ve told them it’s not apples to apples, isn’t a valid test and won’t explain why I’m having the issues I am.

Is there anything I can do on my end to push them into the right direction?