Honestly I am seeing simply randomness here.

If you plot any piece of random data, and then sort by frequency you'll get a graph that looks more or less like this.

I think you should verify your data in two ways:

1: Plot the graph for each year separately, and see how correlated the years are. If the rankings are the same year after year then you have something, otherwise it's just random.

2: A second thing is randomly divide American and non-American each into two groups by model (just randomly assign each model into american 1, american 2, non 1, non 2). Then plot the data for each group in each place, and see if the rankings are the same - if they are, then the American vs not distinction is valid. Otherwise, again, you just have random data.