That doesn’t address the issue at all. The issue is that discarding small changes in any color space is fundamentally lossy, so the algorithm isn’t lossless.
But my comment was addressing parent's point that pure blue and pure red are currently considered close together since they share the same greyscale values.