I think you're missing the point.

The article does not claim that Huffman coding was superseded by dictionary methods. It does not confuse statistical modelling with encoding.