The only problem is that the "image production chain" that produced the images you viewed may not be neither optimal nor showing the strength and weakness of the faveon design.
For example, comparing the jpeg from a canon sensor with that from a Sony sensor you may not notice the striking difference in shadow noise.
I know that it adds that extra degree of variability, but that variability is there in the final print or on-screen image, so it cannot be ignored; that us why the only fair way I see around this is to give the camera maker the opportunity to make its strongest case, by using the raw conversion software and procedures that it provides, or a third party product that it recommends. Tests should of course include difficult scenes like ones of high subject brightness range (so-called high DR), so such comparisons should include whatever resources and guidance the maker provides for handling such scenes, so to assess the shadow noise that you mention.
If the camera maker fails there, it is not much use buying a camera with an allegedly wonderful sensor and raw files that even its maker cannot turn into a good final, viewable product.
To me, what I propose is akin to judging a film primarily by developing and printing according to the maker's instructions, not by imposing some standard development and printing procedure, or somehow examining the _undeveloped_ film. (Raw files are more like undeveloped film than developed negatives or transparencies.)
On the other hand, I can see why some tech. fans who are interested in exploring, comparing, and debating technologies, rather than assessing the actual photographic capabilities provided to them by various tools, might want to spend their time analyzing raw files.
