Is such flag really needed though? As per Guillaume Smet description, implementors of ImplicitNamingStrategy were already facing the weird magic value "collection&&element" in some cases, so it seems just logical to me to make this consistent and do all invocations of that method in the same way.

But if we cannot agree on that, +1 for having this hidden option. Also +1 for finding a saner approach than passing this magic value in the loger term.