I see your point about CR. From what I understand, they rely on their members filling out survey forms every year. This could account for the variation in rating from year to year. I guess you should always be skeptical of a self-selecting sample in a statistical study.
Here's another question. If I average two visits to the shop every year for a repair or recall does this mean the very most reliable cars visit once a year?. With CR it is very difficult to see where the reliabilty of your own car fits into their rating system. Is your experience average ? Better? Worse than their population of survey respondents? Who knows?
Don't get me wrong. I love my Mystique, but it was what I would call a mature design when I bought it in the spring of '97. It seems to have had many reliability issues. It's a pity, because it has many endearing characteristics as well.