I feel the proof is in the results themselves. A few seem off, and I've commented them as such. But this is more a matter of the small sample sizes than the source of the sample.

Here are the results for the BMW 3-Series, in successful repair trips per 100 cars per year, with an * if the sample size is small, and the average odometer in ( ):

2000 200* (119600)
2001 135 (73000)
2002 151* (71500)
2003 104* (56500)
2004 75 (41200)
2005 69* (33900)
2006 68 (24900)
2007 52 (10200)
2008 31 (7300)

Though half of these years have sample sizes below my minimum, we still get a nice clean curve, especially from 2002 on. The repair rate for the 2000 seems a bit high, and that for the 2001 a bit low, but the odometer readings likely account for this--those 2000s have a lot of miles on them. Did the 2000 MY begin early for the E46?