Pixel Size vs Sensor Efficiency

That would be difficult to do. Currently all the columns of one or two rows are transferred in parallel to read-out circuitry above and/or below the sensor array. From there either all columns continue to be processed in parallel, as in Sony's sensors, or they are processed partially serially through several high-speed outputs and A-D converters, like Canon's sensors. Either way the number of parallel first stage amplifiers active at one time in CMOS imaging sensors is normally limited to two times the number of columns in the array. With back illuminated sensors it might be possible to increase the number of rows read in parallel to four or more, since the wiring could completely cover the array without interferring with light gathering, but I haven't heard of sensors doing that yet.