Sorry about the noise, But I have to comment here again about my HB-75P since I had made a mistake, and my above post might be misleading. 
I had mixed composite output and SCART shell ground pins (the cable I have has a detachable pin in the shell). So, by accident, composite out was shorted to ground while I was making the latter testing (and presuming composite disconnected). With that setup, I got an out-of-sync picture with correct colors (with the CVBS signal shorted to ground and pin 19 disconnected) I mentioned above (also, still a blinking pictuere). With the composite pin totally removed (not connected), my TV does not show any picture.
However, if I were OP, I'd still measure the BLANK voltage, since that is easy to do. My rationale for this is, that there are some unofficial extensions (like YPbPr video) to the SCART standard, and it is possible some MSXes (and other equipment) have RGB output that is not adhering to the standard. I come to this conclusion because it seems my HB-75P is one of such devices, since it outputs 5V to pin16 (but it is certainly possible that the HB-75P RGB output is broken, and this is not by design). Considering the original spec has defined anything above 3V, I'd presume some manufaturers might do this - original spec has only low (<0.7V) and high states (1-3v) for the BLANK pin).
I also tested reducing the voltage output my HB-75P by putting a resistor into pin 16 (220ohm) to get the 5V down to 1,7V. Bingo: I still get a blinking picture, but now with correct colours! (but with other artifacts, but I will not go into more details here, since I could aready be blamed about thread hijacking) So there is some improvement here, and now I know my TV does not like RGB signal with 5V at BLANK. But there is something else wrong with my HB-75P's RGB output, too - or it could be just that my TV is picky and does not like sync over composite video, and there's nothing wrong with the HB-75P (except non-standard RGB output) - I have to go to a friends house and test on their TV to make sure.
Although there is a (remote) possibiility this is helpfull for the OP, I will shut up now about my HB-75P in this thread! 
Cheers!
P.s. I still have the NMS-8255 I've been comparing to. The difference is, that it has 3.0V at BLANK (pin 16), and pin 19 has CSYNC (clean sync), and no composite video - composite video is available only via the RCA jacks on it. My TV has no problems with the NMS RGB signal.
Since no one has answered, being a novice, I will - some expert may want to comment if I say something stupid here 
Elrinth: basically, what you say is correct, except always change the multimeter setting to the correct setting before connecting the measuring cables to the live target! Also, never change the setting with the cables connected to the target! (from your post one could understand you are doing the vice versa). With non-powered things, it doesn't matter what you do in this regard.
At voltage measurement, multimeter resistance is high, but for example at current setting, it is low and will short things you try to measure (and no one knows what happens inside the multimeter if you change from OFF->voltage or change voltage range during measuring, so I would not recommend doing that, either, with cables connected to the target).
Be careful when measuring (any live device!), and do not short anything with the multimeter leads! Practice with the cable disconnected (or MSX power switched off). Or, use small "jaw/clamp/crimp connectors" (not sure what they are called in English), if you have some, and put them on the pins before powering on.
Measuring between the SCART casing and BLANK pin should suffice, but to make sure, you may want to use the BLANK ground (pin 18), since that (pin16-pin18) is what the TV will "read" (but the ground should be common, i.e. that pin should be connected to the casing, if the cable is conected to any SCART slot, and I think I've seen some SCART cables with all ground pins connected to each other and the casing, although I think that is not how the cables should be made).
I measured from the SCART cable with the other end connected to the MSX. Someone might consider using the SCART socket directly more safe (with some pins stuck into corresponding holes), since there is only two pins a clumsy multimeter operator might short with the multimeter leads.
