Hold it! Why do you want to filter the Xs? Those are tofu marks that, since there are over 6,000 kanji characters but only 4,096 available spaces in the GRPF sprite sheet, represent those that didn't make the cut. In my screenshot, the green characters are the obscure kanji that got replaced by tofu marks (red). But since they're still REAL kanji, their unicode/CHR$ value remain unique.