Answer from @ATpoint, converted from comment:
A detected gene is one that has a count > 0, and generally the deeper you sequence the more genes you detect until you reach a saturation which is being influenced by the number of biologically-expressed genes and the single-cell technology you're using, e.g. droplet vs plate-based and the library setup, e.g. end-tagged vs full length.
Counts cannot influence the nature of gene translation. A count is an observation; a translated gene sequence is a physical thing. Whether or not a gene can be translated into protein has no relationship to how often it is observed in existence.