Keith Miller explains it well:
The character states used to define higher taxa are determined retrospectively. That is, they are chosen based on a knowledge of the subsequent history of the lineages possessing those traits. They do not reflect the attainment of some objective higher level of morphologic innovation at the time of their appearance. Also, all the features subsequently identified with a particular higher taxon do not appear in a coordinated and simultaneous manner but as character mosaics within numerous closely-related species lineages, many of which are not included in the new higher taxon. In addition, as discussed above, the species associated with the origin and initial radiation of a new taxon are usually not very divergent in morphology. Were it not for the subsequent evolutionary history of the lineages, species spanning the transitions between families, orders, classes, and phyla would be placed in the same lower taxon (Fig. 3).
This is a good point to remember, and as Graham Budd has pointed out, it helps explain why we look back at the Cambrian and have such a hard time classifying a lot of the organisms. Stem group organisms often do not fit comfortably into crown groups.
Exactly. The Cambrian phyla show different body plans, but in each case they represent a very early stage of experimentation with the architecture. They're all rather small and simple compared to the variety of life in the same phyla now.
IOW, they look only recently diverged from each other. (That's using "recently" in the geological sense.)