soapUI performance issue with Layer4 switch

Hello experts.

My customer had a problem with your tool and our L4 switch
In fact, we tried to analyze that issue, our customer didn’t have enough idea with that tool.

Problem description:
The servers use 7003 port for their service and they wanted to use 80 port for their clients.
The topology is soapUI --> L4 (VIP:80) --> servers(Real IP: 7003).
In this case soapUI showed very low tps(around 2 ~ 30 tps). When soapUI tested like this(soapUI --> server(RIP:7003)), there was no problem. tps was around 2 ~ 300 tps.

Although we asked how soapUI worked, they didn’t have any idea. So we changed L4’s Vport to 7003.
So changed topology was soapUI ---- L4 (VIP:7003) -- servers(Real IP: 7003). In this case, soapUI showed good tps (like direct connection).
L4 switch always checks destination port. 80 or 7003 is not issue on L4 switch. If we couldn’t handle soapUI, we should showed bad performance when we used 7003 port on VIP.

We just found two different things from packet capture.
One is:
If the server uses 7003, soapUI inputs port number in Host filed like this: Host: 10.10.20.11:7003.
But the port for soapUI is 80 because L4 switch will translate from 80 to 7003. So the real server will see default host filed like this 10.10.20.11 even though the server uses 7003.
Is that no problem?

Second is:
soapUI just sends two POSTs(HTTP/1.1 and HTTP/0.9) for a test. But there was a big gap(0.14 sec or 0.12 sec) between ACK for first POST and second POST. Please look at packet no.7 and no.8 in Fail-port-2343.pcap.

I will attach two dump packets OK and Fail.
soapUI: 2.5.1
Server: Weblogic 9.x on HP Unix.

Thanks in advance.