raw stills in...converted?

Of course they are converted. Nothing except your camera and dedicated raw converters can work with raw. A raw is just a sensor dump from your camera. It needs to be interpreted by a converter into something an app can recognise as a photo. That's the nature of raw. You're best to get your raws into a raw converter - iPhoto, Lightroom, Aperture whatever - and process them yourself. Then bring that image across to FCP. If you're not processing them yourself, then something else is doing it automatically, and then you might as well shoot jpeg, as your camera will probably do a better job.

When you import images into FCPX, if you leave them at their original location, they stay just like they are. What shows up in the Library is a symbolic link to the original media. They are not converted even when Copied into the Library. This includes CR2, RAW or whatever other format you use. Any conversion occurs in "real time" in the storyline.

When you import any image and have it Optimized, FCPX will convert it to 10 seconds of ProRes 422 video (of the same dimensions as the original). This only makes a difference when you apply it to the storyline. When you apply still images (not optimized) to the storyline, FCPX will use the Preferences setting for Still Images:Editing duration length (so that it is possible to apply single frame images by default.) Otherwise, once optimized, no matter what the setting is for still images, FCPX will apply the "image" as a short clip of fixed time.

The quality of ProRes is regarded as "virtually lossless". You cannot really tell the difference. It should also have the same bit depth (12) as the original RAW images (although somebody with better knowledge about this should chime in here).

The idea of trying to maximize quality is often misunderstood in video. Your raw images are probably 3x4k, might be larger. High def video is 1920x1080 or so. If you use your raw images full frame, you're throwing away huge volumes of pixel content and, unavoidably, some color information in the transcoding from raw to video codecs. As Fox notes, it's visually lossless.

We've been using still images in video production since the days of using a Kodak Carousel to project 35mm slides onto a glass beaded screen and shooting 'em with a three-tube Plumbicon studio camera. Back then, NTSC video was made of 525 scan lines and we could zoom w-a-y in on the projected image but we eventually resolved the glass beads. We also used XYZ animation tables to do the Burning Ken Effect on printed stills long before mr burns was even born.

The 2k and 4k formats available within FCPX (or any other current video tools) are seductive but you can't play them on anything readily available to consumers so there's no real point. Yet.