I had one job that, days before the interview, sent me the spec for an archival format they used, and asked me to implement an extremely basic archive generator and email them my code. At the interview, they asked me to explain what my code did. I went into detail about specific implementation decisions, things that worked for this restricted domain but would not work for a more general version, and ideas for future development.

Best job I ever had.

I think the answer to your question is that it's impossible to determine whether someone is an STE with 15 minutes in front of a whiteboard, because nobody codes like that. You will get false positives from people who can slap together a simple algorithm but don't understand why slapdash hacks aren't good enough for production, and false negatives from people who are brilliant coders but need a little while to get into the headspace, or have their flow disrupted by having to use a damn marker instead of a keyboard and their favorite text editor. You cannot identify an STE in a one-hour interview. If you want to see if someone can code, have them actually code.

Oh, here's another big problem I don't think anyone has mentioned yet: people who think that the stuff they, personally, have memorized represents the core knowledge of computer science that everyone should know, and everything else isn't really important. This is the guy who is unimpressed by absolute mastery of regular expressions, but sneers in disgust when you don't know all the arguments for pushd off the top of your head.