Fine, here you have the definition of a transaction unit. Of course I understand that there are multiple inputs/outputs possible, which makes bitcoin transfers much more efficient than in conventional payment systems. I was just using terminology common on sites such as blockchain.info. This makes 4 tx/s a valid description of the real-world situation.

It is no good having theoretical efficiency available if it is not being physically used! You say it could be used, and this is true, however optimum transaction efficiency cannot be achieved by a wide user-base overnight. My concern all along is that bitcoin usage is growing exponentially, at a much faster rate than the adoption of transaction efficiency, or complimentary off-chain solutions are appearing.

There are a lot of smart people on this forum who fully understand exponential growth curves. Yet it seems some of these people think that a hard limit can seamlessly force adoption of efficiencies and complementary solutions which are developing linearly in comparison. 

There can be no seamless transition in that scenario, which is why an adaptive/algorithmic block size limit is needed sooner than later.