No, you do NOT have to have the same number of observations between one brand and another to make valid comparisons between them. Look up "levels of signifcance."
Having "actual data" is exactly what DOES make Consumer Reports' survey important. Is it perfect? No. But it's much more valid than, say, a self-report online poll where the general public is invited to vote their opinions. That has no statistical validity because it's just a popularity-contest election. CR, at least, has its audience and surveys it about the readers' OWN cars. The PERCENTAGE of people who report problems with each car brand are the basis of their numbers, and if those numbers are large enough, the results are probably meaningful. To CR's credit, when they don't have enough numbers, you'll see in the magazine's charts that they simply don't publish a reilability result for that model; they say "Insufficient data."
Speaking of a sample size that's too small, every person in this thread who's said "Consumer Reports' survey about Brand X is all wrong because my 2016 Brand X has been perfect" is drawing trend conclusions based on a sample size of 1 vehicle.
