I didn't say sorting is never done by stroke count alone. But I have seen radical+residual stroke count much more often than stroke count alone. Probably a result of the content I'm accessing. It's mostly Japanese and not intended for children.
The dictionary and non-dictionary sorting distinction that you make doesn't sound like a real thing. The audience, the country, and the number of items sorted are bigger factors. But you're not wrong in that stroke count is sometimes used alone.