There are no real conclusions that you can draw from 100 names. There's no way to know if the most common starting letters are A, C, K and S across TER as a whole based on what you've analyzed. I looked at a few common names in DC, and there are about 120 Nikkis, but that's out of something over 12,000 providers listed in DC. So even though there are (have been) a bunch of Nikkis in this area, that's only 1 percent, so if you randomly picked 100 providers from the database here you'd be lucky to get even 1 Nikki in your sample. But there are only 15 Kellis listed in DC, which may or may not be representative of TER as a whole, so it kind of surprises me that you got 4 in your sample of 100. It probably points to some sort of unintended bias in your selection criteria. You said that you chose based on a subset defined by age, body, hair; for some reason when I see the name Kelli (spelled that way) it brings to mind ads I've seen for young, blonde spinners. Any chance that this matches your criteria? Shot in the dark.  

As to your second question, why or how they pick their name. There's no standard approach, the reasons are going to be as varied as the providers themselves. There will probably be streaks of relatively popular names, but it will almost certainly be in small clusters relative to the size of TER overall.  

And the final question, yeah, there are going to be some names that you would think would be bad or at least odd choices, but in a data population the size of TER you're going to see all sorts. There are 3 ladies named Norma in the DC listings, that's just not a name that gets my engine running