This is the expected, yet not quite documented behaviour. We cannot load the first N particles from a CSV file because we do not know how many particles are in the file.
We made the decision to avoid counting the number of newline characters in the file because this is non-trivial for any decent sized particle file.