Basically, if the details of a quest can change depending on the situation, how can they possibly voice every single possibility without inserting words in robotically and making the quest giver sound like an automated tech-support line?
My personal hope? They simply choose not to voice some things. Let us read for the drudgery of Enter Cave A, Kill Foozle, Retrieve Thingamabob. Let's hear the voices for the important stuff. Non-voiced mod work would suddenly blend much easier with the rest of the game.