…are weighted against national averages and plot the most distinctive pick for each state—the brand or whatever where a state has the largest deviation from the average.

So if Bud was 60% of sales nationally, if Bud was 60% of sales in a state it wouldn’t get flagged since it’s no more popular there than anywhere else. So instead it would tag a brand that was 10% of the national market but 15% of the local market, since that had the biggest local jump.

Not sure if that’s how this map was generated, but something like that could explain odd results,