Good question Greg C, and one that I have also previously pondered!
Obviously, if every exit door in a classroom to the open air is considered a proper escape route for persons anywhere in the building, then in an L4 system, automatic detection would also be required in the classroom. And if the classroom is big, then more than one might be required.
My previous pondering has led me to the following conclusion (rightly or wrongly!);
The classroom is not part of the common escape route unless signage from the corridor/circulation area indicates that the route is an emergency exit.
In my view, the exit doors to the open air from classrooms are primarily planned for the use only of those persons in that classroom i.e so they don't need to enter the corridor in an emergency situation (especially where they can't in an emergency). These classroom doors invariably do not have any proper emergency exit door hardware or, are normally not as wide as proper emergency exit doors. This is probably why signage doesn't normally direct persons in a corridor/circulation area to use a route through a classroom as a means of emergency exit.
If my conclusion is right, then no detection is required in the classroom for an L4 system unless signage outside the room indicates that the route through the classroom is an emergency exit route.
The second part of your question I believe could be argued as being fully answered by the recommendations of BS which allows no variation in the amount of detection required in any area i.e. if it has to have detection the recommendations for detection quantity/spacing etc. always apply.
However, Lord Toddy always insists that system designers should be brave and use their knowledge, experience and commonsense in most situations and not always slavishly follow a BS recommendation when it's application in a certain situation seems to be 'over the top'
Applying this argument to your specific question, if you had a room that was, say, 6m wide and 50m long (silly dimensions I accept!) and you had an exit route that crossed the width only and totally at one end of the room, and you decided to cover this route with automatic detection in a L4 system then surely one detector only at the end of the room with the exit route would be sufficient to detect the amount of smoke that might affect the safe use of that exit route? A 'codehugger' certainly wouldn't agree with this argument but I see this being an example of Mr Todd's suggested way of looking at some things.
To further support both the above conclusions, I would argue that providing automatic smoke detection doesn't actually stop anyone from trying to use an exit route that is affected by smoke in that area. People could still take a route that led them to where the smoke exists! The detection in an L4 system is installed to provide an early warning that there is sufficient smoke somewhere on an exit route to operate a detector, and that if immediate evacuation is not commenced these routes may become filled with enough smoke to eventually prevent their use as a means of escape.