Simulation methods, such as the one employed above, are used to test statistical techniques, to assess their accuracy. Presently, the few viable options include N-body simulations, which are a prohibitively slow to produce thousands of realizations, or the faster LPT2 methods, which cannot reproduce small scale clustering. The above method is really fast, and was designed to simulate small scales relatively accurately. In the near future it will be merged with other tools, such as LPT2, or Zeldovich approximation, to include large scale clustering as well. If you guessed correctly, you probably looked at the largest structures on the figures (or, you had shear luck).
The answer is: upper right corner. Presently I have no cgi scripts installed to record and log correct answers. I might do that later (but don't hold your breath)