> Last voter data I saw has registered Republican's in the low teens as a percentage of registered voters.

Not that high; per current registration data [1], 37,719 registered Republicans of 437,529 total registered voters, so Republicans are 8.62%. Low teens is about what the typical Republican running in San Francisco gets on election day

> The only way I could see a Republican winning SF is if he/she were a fiscal conservative/social libertarian.

If the political mood in San Francisco was running in a way that social libertarian positions were strong enough that they'd get enough liberal voters to hold their breath and vote for a fiscal conservative, then a left-libertarian would probably end up as the Democratic candidate in the general election, and win. (Although the main effect, given the way California's "open primary" works, would likely be a contested primary among Democrats resulting in the two candidates that make it to the general election both being Democrats.)

I suppose, though, if you the the surge of social libertarianism happens after the primary but before the general election, and the Democratic candidate isn't a social libertarian, you might end up with a chance for the Republican. But that's a pretty freakish scenario.

[1] http://www.sfelections.org/tools/election_data/

Also note that's "the typical (republican running in San Francisco)", as distinct from "(the typical republican) running in San Francisco", who would likely get even fewer votes.