Playing with the y-axis in this case, doesn't change the shape of the data. It only makes the graph more readable. For example, would you start the y-axis on a graph of global temperature at –273.15°C (absolute zero)?
Second, why should the graph use evenly sized groups? The whole point is to show an outlier in tax data among a small group of people. Using evenly sized groups obscures this point. If a graph of global temperature only displayed in increments of 1000 years, it would likewise obscure the recent impact of human industry on temperature.
Finally, as the article notes, the graph is of actual taxes paid not historical taxes.
If there is any legitimate criticism, it's that the graph probably should have been a bar graph instead so the "top 400" (likely for ergonomics) is less jarring. Also, just saying that the numbers are known for inaccuracy? This is a lazy statement that needs to be backed up.