This raises a question, though:

Why does the CPS do things this way?

Why don't they adjust historical numbers to match the yearly revisions? Why do they rigidly follow the Census' population adjustments that can distort statistics like you said? Why don't they use a larger sample size?