I misread the TI doc. What they actually said was crud getting in the mirror caused about 68% of the failures. They identified and implemented design changes to fix the vast majority of the other issues in the first year of production, including things like tiny springs on the mirror edges and smarter electronics to prevent hinge memory. The only significant remaining cause of failure was contamination by particles during fabrication and assembly. I mis-read that to mean fab at TI and assembly by the OEM, but after Iggy posted I reread the article and from the context it's clear that they are in fact speaking of crud inside the package. It's a very interesting read. Apparently the dirt comes from many sources, including various steps in the fab and machining process.

As AzRyan noted, the higher res and smaller the screen (basically the smaller each pixel was) the less the mirror artifacts were visible. Also it looked as if the lens (screen) on RP DLPs might be helping to soften the lines and squares. The projector I saw that was really poor IMO was a 800x600 Insight (IIRC) on a really big screen. This was one of my first looks at a DLP system and after that I was looking for the pixels.