I figured the reason GTool failed must be that the new PC was faster. I started disassembling the tsf3 code and looking at it (since it was smaller than Gtool). I tried putting in huge delays in the port I/O routines and still it would fail. Time to really look at what it was trying to do. It is using a somewhat arcane semi-undocumented feature of the old IBM parallel port hardware. It tries to read some kind of status from the GD off of the pport control lines which are normally output only.
So what it boils down to is that almost all ECP parallel ports produced today change the orignal IBM pport hardware to support higher transfer speeds lowering the value of the pull-up resistors on the control port lines and some even eliminate the control port readback path (probably since it doesn't work for squat anyway once they changed the pull-ups).
I had tried tsf3 on two computers (two different parallel ports) and it failed and I had tried an add on ECP parallel port card without success. Today I have finally proved that the problem is in the parallel port. I have found an add in parallel port card that works, even on my fast PC. That is the answer: your parallel port needs to be compatable with tsf3! If it isn't it won't work. If it is, you're golden.
The card I got to work is a Lava Parallel-PCI EPP card (made by www.lavalink.com). It may be because it's EPP only and not ECP, but whatever the reason, it works. I recommend anyone who hasn't been able to get tsf3 to work rush out and buy this card. Anyone who thinks they'll ever upgrade their PC and will want to run tsf3 on the new one I'd recommend you pick one up too because the odds are tsf3 isn't going to work on a new PC's parallel port so be prepared.