Take the following statement with a grain of salt, considering how long it's been since my car's had SUs on it, but I too feel that the spring should be compressed while setting the float height.
The spring's tension is minuscule, IIRC, and the buoyancy of the float in a pool of fuel would be enough to compress it, raising the level of fuel in the jet well. I suppose it is theoretically possible to *tune* for the difference, by lowering the jet height, but that would leave less reserve in the float bowl(s), and engine smoothness under severe braking, acceleration and cornering may be compromised.
Pure speculation, of course!
For a comparison, Weber floats are set hanging vertically__pivot at the top__and measured WITH the gasket in place and the spring not compressed. Again, it takes very little to compress the spring-loaded pin in the inlet jet assembly (available in a variety of flow capacities) so that must've been taken into consideration when specifying the recommended float heights.
A final note on setting the float heights with multiple carbs: it is far more important that the float heights be equal for all units, than to worry if they're a touch high or low. Instantaneous throttle response depends on them all
picking up at the same time, and the throttle-plate synchronization and float heights dictate this for optimum results. If the pairs__or trios__of cylinders had separate cams and crankshafts, then synchronization wouldn't be such a big deal