I didn't say _perceptual_ watermarking, which, as you point out, is at odds with perceptual audio compression... :)