The reason that the the sequence is 121 and not 111 is because the designers of this specific implementation wanted to give a greater weight (2) to pixels in the same row/column as the pixel in question.

Does a greater the magnitude of the weights means a pixel will be brighter or more saturated? And is there a upper limit to the magnitude of the weights?

@Claire I think in this case, the weights are multipliers for RGB values, so greater magnitude would mean a pixel becomes brighter and there wouldn't be an upper limit besides what would yield reasonable results (too high, everything becomes white)