Other sources? I used this site to test for quality of the G70 prior to purchase. I think many sites such as this one provide that opportunity for potential buyers. The owner experience is real and personal and unfiltered. An experienced reader can discount those few posts that are rants and unrealistic. The rest provide a distilled opinion of hundreds of drivers. I chose that as more authoritative than the CR review when I bought my G70. Though the CR review was generally positive and the negatives were issues less important to me. The full review was posted here by someone, text and all.
There is also TrueDelta, based on user surveys, and the person who runs it is a statistician with a doctorate, so you could imagine they know what they are doing, and I have them to be kind of accurate? If you look, you'll see that their results generally tend to correlate with CR, so that's a good sign that they are both on the right track - when they have enough data.
JDpower is complete BS, though. They are an industry tool, designed to drive business and act as marketing.
Repair data is literally impossible to get as a direct measure. That would require a network of databases/data from hundreds of different parent sources, all of which could be considered to be in direct competition to each other (in one form or another) and that data would be proprietary as hell.
However, back in my youth I worked for RP Alpha group, which is a large advertising and marketing agency out of California, and our largest account was American Honda. While we performed tons of original research with dealers, customers, potential customers, etc.. there was also lots of industry-specific aggregate data for the industry and for marketing - Allison-Fischer, folks like that. Everything under the sun was available for us to use, every bit of demographic, descriptive, reporting, federal, state, foreign, - everything. Everything, except repair or actual quality of product data. That just doesn't exist outside of the organizations that own it, and /or choose to share it.
So the absolute only reasonable way for anyone to get data (as far as I am aware, perhaps something has changed since I was in the business) on vehicle quality, perception, etc., is to perform surveys and ask people. It's not hard, there are many ways to do it. If you ask enough people the same questions, you can be pretty assured that results are accurate. Statistics is branch of mathematics and as such is objective and by definition unbiased. Poor data is easily identified and removed. Results and data go through many, many tests, and are generally accurate. Businesses, governments, militaries and police use these methods every day to make life and death decisions. So it's good enough for the us military, but not for the average car guy just wanting to see if that 2-year old M3 he is thinking of buying will more likely than not shit the bed in some way before ownership is out? Additionally, these kinds of stats just give likelihood of x or y, sometimes more, sometimes less, but it's never 'oh yes, for sure', or 'oh my gosh, never.'
I think it's more like people generally don't want how they *feel* about something contradicted by the results of a process they possibly don't understand and are suspicious of. Especially today. People are told by advertisers, businesses and politicians, facebook buddies, dudes at the bar, how they 'feel' is the most important thing, and that data, results, accurate predictions, accurate warnings are all somebody wanting to control them, or lies to prevent them from doing what they 'feel' which of course, must be correct because it's how they 'feel.' Which to me, is sort of like the Brawndo argument, which sort of sums up today's world anyway.