Just to circle back to the whole Benford's law thing and why it is inapplicable, let's look at some good examples of where it is applicable and see if that helps understand why it is not here. A perfect example is street addresses, which is not only a famous example, but one that is very illuminating. Imagine all the streets in the country. We have short streets, long streets etc. For each of those streets, they start somewhere around 1 or 100, and go up. Now, let's let our streets start at 100, then for the first 100 street numbers, 1 will be the first digit. It will be 50% of the first digits for the first two hundred, 33% for the first 300 etc until it is 11% after the first 900 numbers, taking us to 1000. After the next thousand numbers, 1 will be the first number for 1100 of the first 2000 street numbers, then decreasing, then increasing etc. The catch, which makes the law work, is that there are many streets that are in the low numbers, and fewer in the high numbers, so there is always a better chance of 1 and 2 being overrepresented. I am going to assume this is clear, ask if it is not.
Now take artificial groups of 1000 people known as voting precincts. Look at our previous example and you see that, absent voting preferences, the probablility of a 1 being first is 11% as it is for all the other numbers, simply because of the artificial size of the precincts. Adding in voter preferences in a two party system doesn't do anything to restore it to Benfords law, which does not hold because Benford's law is specifically a law of large random numbers, while a precinct style voting system with two options is neither a large nor a random number, which means there is no reason to expect it to conform. It has none of the sequential properties above that would suggest a Benford distribution. Rather, we would expect a distribution that conforms to partisan preferences overlaid on populations of 4-700 people, representing those who actually vote, which is kind of what we got.