Sure: Windows has a lot of internal "code fat," in addition to having to run necessary applications like antivirus, that take up CPU/system resources and slow down the broadband conneciton.
I couldn't compare XP on that system because it won't even run on a lowly K-7.
As a quick test, I just ran what I think is the same speed measurement test you ran (speakeasy.net) on four configurations: my main laptop, which is a P4 Centrino 1.7GHz with 2GB of 533MHz DDR2 on XP Pro SP2; an Ubuntu browser appliance running on VMWare Player under XP on that same system; my backup laptop, which is an AMD Athlon 1.8GHz with 512MB of 400MHz DDR, also on XP SP2, and the same backup laptop booted to a PCLinuxOS 0.93 LiveCD. The Ubuntu virtual machine and PCLinuxOS LiveCD had no applications running other than the browsers (Firefox and Konqueror(?), respectively); I did not bother to shut down any applications on either laptop running XP. I performed ten speed tests to each of three of the possible testing servers under each configuration.
As expected, I noted no statistically significant difference among any of the four tested configurations for either upstream or downstream speed. On the newer laptop, Firefox never used more than 2% of system resources during the tests in XP. On the older laptop, Firefox peaked at about 30% of system resources in testing. I don't know how to monitor such things under Linux but I would expect similar results.
I respectfully submit that the large disparity you noted was either an anomaly in the single test or a problem with your Win 98 setup. I admit that I haven't used Win 98 in about five years, nor do I recommend it, but I do not believe it will process a 10MB data stream at less than 4MB. Linux may have certain speed advantages in certain situations, but at broadband speeds I do not think they come into play.