I noticed that the mas5 in affy command seems to include control genes (genes that begin with AFFX-) when scaling data. I believe Affymetrix excludes these genes. Perhaps this could account for the part of difference in values between Affymetrix's MAS5 and Bioconductor's?