Thanks for your comments, guys.
I've only tried this 4 times - the first time I got an overexposed blob, the second time I got a whole avi of black frames with not a single hit recorded, and the third and fourth times are the results above.
It's essential to get the finderscope as accurately aligned as possible, get the focus spot on, and make a rough guess at the exposure. I usually put the 12" newt on my EQ mount to get focus right (so the star doesn't move) and rough adjustment on the finderscope, then move it to the dob base to align the finderscope again. I've found that looking through the finderscope even on a slightly different angle can mean hitting or missing the ISS in the FOV.
I guess at the exposure, starting at around 1/1200s for a bright overhead pass. I'm guessing it would need to be a bit less for dimmer passes.
For processing, I go through the avi in virtual dub, deleting the blank or blurry frames. When all that's left is frames with bits of the ISS in them, I save them as bitmaps. I then use ninox/ppmcentre to centre the ISS in the frame and crop the frame to something like 400x400.
I then hand-select the sharpest, best frames and layer (stack) them in photoshop, then save as tiff, take into registax for slight wavelet sharpening.
For the second shot above, I was lucky enough to capture a few frames where the ISS went behind a tree - this reduced the exposure and the overexposed areas (eg: the radiator on the right) came out at about the right exposure. So I processed them independantly, and used a layer mask to combine it with the overexposed one.
I look forward to trying again in a few days with another bright overhead pass - but the weather doesn't look good!
