For NTSC operation, link S28 can be changed to give a constant
reference phase (IC83 pin 13 to 0v) - R92 must be removed for NTSC,
and the crystal X2 must be 4 times the colour carrier frequency of the
NTSC broadcast standard (eg 14.318MHz for USA).[/i]
A few evening hours of probing and circuit tracing later and it turns out that the above is only partly true for converting a later NTSC board to PAL, rather than an earlier PAL board to NTSC.
It turns out that there are only 2 export models of the BBC model B - West Germany and the US. They both used the exact same PCB layout, which is what I have; namely board revision number 223.000. Mine is Issue 1, though I cannot find any information indicating that there may have been other issues of this board layout.
The PAL/NTSC colour burst reference on this board is jumper S38, rather than S28. There is no resistor to either remove or replace to convert between PAL and NTSC.
After changing the position of J38 from the NTSC position to the PAL position, I met with success, still injecting the correct chrominance sub-carrier frequency from my Rigol DG1022. I wrote a quick basic program as a colour test, which produced the screen below.
However, so far, the colour lock is only reliable and stable so long as I have the probe for my DS1202CA oscilloscope attached to a particular node of the chrominance modulation resistor matrix! I suspect that this is because the DG1022 signal generator output isn't clean enough for the job - a little probe capacitance in the right spot delivers just the right amount of HF filtering. When my 17.73 MHz crystal arrives from Mouser I report on the results.
I would still like to find a schematic for my actual PCB though, but so far the web has not turned anything up.