Post camera processing can cram in as much additional meta data as desired, so I assume you are interested in what the camera produces directly.
In general, RAW is going to have everything the camera offers. Conversion to another format isn't going to create more meta data. It's certainly possible that a given manufacturer may only insert selected meta data in another post process format, but I'm not aware of it with one exception. Multi-image processing in-camera, such as Panorama and HDR may include additional information relative to the additional processing.
WebP has some nice features, but none of those apply to your stated purpose of computing MTF. You already stated lossy compression is undesired, and transparency support would not be used. Most importantly, the output is 8-bits per color channel, loosing dynamic range available in the raw.
For out of camera MTF computation, you need to process the RAW or a format that handles extended dynamic range such as TIFF-16/32/float. In either case you would want to keep the original RAW and its meta data for potential re-processing.