The basic philosophy of the distance scale ladder is to use one distance measuring technique to calibrate another that can be used to larger distances. All systematic errors associated with each step are cumulative in the final result! This sucks but we are stuck with it.
The parallax horizon is about 25 pc (not very big). But this region does contain a representative sample of low mass stars.
Cepheids are pulsationally unstable and hence change their radius in a periodic manner. As the radius changes so does the luminosity of the star. The rate of change of the radius is correlated with the luminosty through some real physics!
Hence the distance can be determined. In the case of the Hyades and Pleidaes clusters, they are young and still have thier upper main sequence stars as well. These stars are generally missing from the nearby parallax sample. Since the upper main sequence is significantly brighter than the lower main sequence, those stars can be detected in more distant clusters, where the lower main sequence stars are too faint to detect. It is these more distant clusters that contain Cepheids.