Whoops, now I understand why it seemed to have such a weird spectra.

On the JPL raw images, they have the filter wavelengths and filter IDs swapped right/left.

That is, the image captions will read something like, "Right filter 6 (480nm)"

Unfortunately, the captions consistently show the left filter wavelengths for the right-hand filters, and vice versa. In my example, it really should read, "Right filter 6 (930nm)"

I just used the wavelengths as specified, and thus got it wrong -- I was using near IR for blue and green, and blue-green for Near IR.

Doug is therefore right, the crater rim is very bright in near IR. Goes from very bright to very dark from about 700 to 600nm. That also makes it consistent with the outcrop, which shows a similar pattern if I recall.

For my own reference: RIght = IR, left = visible