Microarray vs Proteomics

At this point, I believe we have a consensus:  Quantification data from microarrays doesn't replicate very well at the protein level.  The problem is that we don't seem to have an idea how bad the discrepancy is or why this occurs.

Years ago, Dr. Rich Helm, who runs the Mass Spectrometry Consortium at Virginia Tech said that in bacteria, the correlation is less than 0.20.  So, 1 in 5 observations match between the two in an organism that has 1 chromosome and very limited post-translational modifications?

I throw this out there because I recently participated in a study where we did quantitative proteomics and microarrays on two cell lines from mice.  Our correlation?  0.01.  One in one hundred observations matched between the two.

The ridiculous part?  That the two techniques gave us extremely complementary data.  They pointed at exactly the same pathway, with almost no overlap in proteins.  As we construct this paper, I'm sure that I'll continue to search the literature.  Maybe someone has a good explanation, but I've been doing some reading and I sure haven't come upon one yet.