Well yes this is what I used to think too and you are right for the coil spring clutch use. But the later diaphragm clutch system is actually quite clever and does not use a spring.
The short slave cylinder e.g C21470 uses a return spring so that the slave piston bottoms in the bore when the clutch pedal is released. Under this condition you do indeed adjust the push rod to achieve a slight clearance at the throw out bearing.
But with a long slave cylinder e.g. C24145 the piston takes a position well away from the bottom of the cylinder bore and so a spring return would cause great free play at the thrust bearing. I think that the head of fluid in the hydraulic circuit is sufficient to apply just enough force on the operating rod to keep it in contact with the operating fork and thus light contact at the throw-out bearing face. This design means that you do not need to adjust the operating rod length as the clutch wears whereas with the short cylinder, as the clutch wears the clearance is taken up and requires a reset.
The rule seems to be that with a spring clutch you use a short slave cylinder plus spring. The diaphragm clutch is used with a long slave and no spring.
Refer also to the bottom of page 116 of the Mark 2 parts manual.