Why examples first
The microformats process requirement to first document examples of real world publishing behavior is an adaptation of the "gather information" step in the scientific method to standards development. The microformats process is the first to explicitly adopt and require this methodology among standards efforts, and for a while was the only effort to do so in the world of standards development as compared to other standards development organizations. The whatwg has also adopted a policy of requesting real world examples as a basis for additions, changes, etc., and cited the microformats process as inspiration for this method.
Common behavior in typical standards development organizations is to simply invent standards / formats based upon intuition or so-called "expertise", and to either not bother collecting examples at all, or to do so only incidentally, or for brief illustration, always serving at best as secondary. Sometimes this intuition masquerades as formal "requirements" that are similarly invented rather than based upon actual publishing practice in the real world.
That non-scientific technique has been tried in many (most) standards and results more often than not in bloated overly complex (certainly not "micro") standards. There are exceptions, where an individual with exceptional discipline and near obsession with simplicity makes something small and elegant, but they are the exception, not the rule, and even in those cases you will often find features that no one ever ends up using.
This particular contrast between traditional standards development and microformats can also be summarized as the philosophical difference between a priori and a posteriori knowledge. That is, traditional standards take more of an a priori approach, independent of real-world experience (data), while microformats take an a posteriori approach that is dependent on real-world experience, where "experience" in this sentence refers to actual "experiences" that occur in the real-world, in particular the Web, and are documented as such, not someone being "experienced" or an expert in their field etc.
What about hCard and hCalendar
But what about hCard and hCalendar? Both of those seem to have been created from existing formats rather than using research of real-world publishing examples and schemas implied therein. What happened there?
In short, hCard and hCalendar were developed before the process was written-up and in fact helped inspire both the "research previous formats" step in the process, and the call to document real world examples - from a lesson learned perspective.
References and see also: