KEGG filters at the MAP or MODULE level, but it merely highlights the enzymes of the mosaic MAP/MODULE that it inferred is present in the organism, rather than display the subset of the network that is actually present.

I agree with you regarding their mutual incompleteness. There's plenty more literature to curate! See my systematic omparison of KEGG and MetaCyc for more details:

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3665663/