And here is more from one of the experienced members of the FTDNA forum. I'll summarize what I think all of their inputs mean in another post:
In the case of most calculators, the populations are used to represent certain global areas, like European, and this is where PCA comes in. Principle component analysis looks at the data and statistically determines how much "strength" , kind of like a heat map, a person has in comparison to a dataset, population set, or continental representation. It would almost be compared to placing a person in between a bunch of evenly spaced samples and seeing which ones have the most gravitation pull.
A person's admixture is completely unique to that person and is based on thousands of years of past ancestry reflected in the comparison to the population sets. Recent timeframe mixtures, say two people from different races, simply means there are two different "thousands of years of ancestry" sets in the comparison mixed together.
The goal for choosing a certain calculator is to choose the one the most closely reflects the ancestry of the person being tested. Some are globally generic, some are more specific to a certain set of populations.