There are not many good sources of reliability data. People like to complain about the few sources we have, but the alternative is relying mostly on internet gossip.
Consumer's Reports is real data, I've yet to see something better. Yes, one must be a subscriber, which means they probably buy the vehicles CR recommends, or why would they waste their money/time to subscribe? The sample size of the "other" vehicles may be very small. Additionally, people who are motivated to report are the ones reporting and may not be a representative sample. However, is the motivation of the Fiat reporter different from the Toyota reporter? IMO you need to also look at the problem areas which make up the 1-5 scale to understand the value of the rating. Most antidotal evidence seems to suggest the higher rated brands often perform later in life as well. CR is not perfect measure, but IMO the data generally has a strong basis in reality.
JD Power "initial quality" - Most of this has to do with knobs, rattles, infotainment, rather than things which matter most. This sort of thing can fluctuate greatly, which is why you see wild swings in the data. Their dependability rating seem like the more valuable one.
