Yes, it picks a blend, but the blend will look like the richest cylinder. The exception might be a misfire where gobbs of O2 aren't burned. This can make the whole mixture look lean. (To the O2 sensor... not an exhaust gas analyzer)
If the car is running at .5 to .8% CO as a Lambda system will try to achieve. The CO levels are read from the rich cylinders as lean ones have none (virtually). A cylinder or two running at .1% CO makes very little difference to the average in this type of testing.
The controlled mixture doesn't compensate for single cylinders and with mixtures as lean as they are a 10% difference in injector flow will be felt at idle. These situations can be determined by the test I described. The test proves that mixture is a problem. It doesn't prove whether it is a difference cyl to cyl or whether the whole motor won't run at lambda control mixture.
It is hard to test individual cylinders without a differential flow test, which will cost as much as a few injectors and probably can't be done in most cases for lack of equipment.
Bosch Master, ASE Master, L1
33 years MB technician