You could always place a "Runs best on Firefox" icon on the portal. (Hey, I have to constantly curse narrow-minded web site designers who think that everyone uses IE and who therefore don't even try to make their site usable with Firefox - it's about time there was some payback.)
But more seriously, here are some of the things I would try in this situation.
1) Run a proxy server and route every request through that proxy server. Both JMeter and Grinder come with proxy servers. The nice thing is that the proxy servers capture all of the web requests, mainly so that you can replay it. But those requests include timings that can tell you what the delay was between the requests. (I think the timings are on by default in Grinder, but have to be selected in JMeter.). Once the proxy is set up, access the portal with Firefox, and then again with IE. Compare the timings as captured by the proxy. That might give you some idea of where to start looking.
2) Compare the html source of the final page as grabbed by IE and Firefox. This might also give some hints as to what is different.