Later edits:
I guess I misunderstood your risetime range. You meant the time between real rise and expected rise, right? The digital signal cannot be that slow. Then what signal did you use to trigger the oscilloscope?
I vaguely recalled you compared the timing of serial, serialport and IOPort. Now I dug the old message out!
The different delay and variation are easier to understand. Did you show the signal of TXD pin at the oscilloscope, and measure the risetime? I would expect the risetime is irrelevant to software, but is a property of hardware. Apparently I missed something.
Thanks.
-Xiangrui