According to the textbook, the problem with "weird" samples is that they are NOT:
very easy to measure.
good for theory testing mode.
very representative.
used frequently enough.