No, a lexical analyzer does not return a token for a comment. It recognizes an comment and then restarts, getting a token after the comment, without yielding any token for the comment.

That way, the parser does not need to be concerned with comments. That is important since a comment might conceivably occur between any two tokens.