Measuring CO Emissions and the NCT

We took Lise's Opel Corsa Champion for its bienniel checkup on Friday; it's just coming up to its sixth birthday. The car has been a grave disappointment to her, presenting a litany of troubles over the years. For instance:

• The front passenger seatbelt catch failed; eventually, they had a recall on the whole fleet about that one.
• There were a considerable number of flimsy, or flimsily-attached pieces of plastic around the seat area that came off within a short time.
• Several dashboard lights failed, and it turns out that they are so awkward to get at that it costs 100 euros labour to replace them.
• The thing has an idle motor (Why? Can't the ordinary motor idle?). This failed just after the warranty ran out, and again three years later. 300 euro a pop. When it fails, it's a complete shambles; you're there, revving the engine in neutral when you have to stop in traffic, to avoid it cutting out.
• We've replaced the exhaust (US: muffler) five times, no less.
• Last test it failed on the headlight aim, and it took major surgery to get the thing pointing where it should.
• The exhaust came apart once in the middle, and the little car vaulted over the back box, folding over the pipe and jamming the box between the floor and the road to make a first-rate brake. Lise was stuck fast. When I arrived down to see what was wrong, I had to cast about for a log, and jam it under the wheel so the car could drive up on the log and free the exhaust. Driving home without the exhaust was the only time the car really went!
• It often produces an appalling rattle, a bit like one of those old Leyland buses starting up, as you start off.

So we approached the National Car Test with no great hopes, and were pleasantly surprised when they told us that the only thing wrong was the carbon monoxide emissions at low idle, and that this would probably be fine if we got the holes in the exhaust fixed. Never mind that the exhaust was practically new, and we knew nothing about holes. Holes can be fixed, and if not we could always buy a sixth set. I'd been fully prepared to write the thing off and take her back to Toyota.

I was a bit shocked at the reading: 3% CO, but the tester said that with holes you get a false reading.

Sure enough, it turned out that the leaks were at the junctions, and were due to the heroic efforts of our usual mechanic to strap up the exhaust and stop the awful rattle. And as soon as the junctions were repaired, the CO reading came down to something tiny, and the car passed.

The main reason for this post is to ask a question that bothers me: How is that holes in the exhaust raise the CO concentration reading? What is that NCT machine doing? It's supposed to measure CO percentage by volume. As far as I could see, it was attached to the end of the pipe. Was the CO percentage different between the two tests, or is the machine measuring some proxy for the concentration?