It's nice to say always enter a full, place name; but many of us started genealogy when we were keeping notes in spiral binders. When we converted to digital records, we had to deal with what we had. Also, some records simply won't give the whole place name and you are left in a quandry.
I do what pretty much what a previous poster does.
If a county only, I always add "County" to be sure to distinguish it from a town or city or hamlet, etc.
If I have place that doesn't give a county and I can't find it in a county finder or Wikipedia, I just leave it the way it was in the record.
I normally try to put "twp" (which is accepted modifier to a place name that will not interfere with resolving a name) for townships to distinguish from other place names in that county.
If the record gives me a name that is not in PNA or Wikipedia and is probably a historical place name, or the place had different boundaries at the time of the record, I don't resolve those place names - I just leave them alone. This means that the heirarchical view will be incomplete, which is why I don't use the heirarchical view much as it only includs resolved names.
I am still plagued by the "resolve" feature not handling five part place names. There are some place names that are both a hamlet and a village in a New York County. Or a division of the City of Los Angeles and also an unincorporated "place" in the County of Los Angeles.