the problem with not being able to load conversational grammar (.cfg) files on Windows 7 is related to the main issue with /eva to begin with. the problem is i've had to manually generate n-grams using the conversational grammar builder in Speech Server 2007. instead, this process should be automated so the n-grams can be recreated nightly or as new music files are added.
Nick (MS) tipped me off to the Dictation Resource Kit. its not pretty, but it looks like it will work for generating the grammars that i need. and it'll take some grunt work, but i think it can be automated. finally, i would have to create the grammars differently in code, and that might mean they would work in Windows 7. the one ugly part is that the grammars have to be registered ... and i'd rather my app didnt touch the registry to avoid the security hassle.