A radiator's catalogue output is not a property of the radiator. It is a measurement taken at one stated condition, and away from that condition it is a work of fiction.

$\dot{Q} = \dot{Q}_r \left(\dfrac{\Delta T}{\Delta T_r}\right)^{n}$ — read aloud Q-dot equals Q-rated times the ratio of delta-T to delta-T-rated, to the power n. $\dot{Q}$ is the actual output in watts; $\dot{Q}_r$ is the catalogue rated output, also in watts; $\Delta T$ is the actual water-to-air temperature difference — mean water temperature minus room air, in kelvins; $\Delta T_r$ is the rated version of that same difference, 50 K under EN 442; and $n$ is the emitter exponent, a bare number the manufacturer measures. The subscript r means rated everywhere it appears — top of the ratio is your building, bottom is the test lab.

That exponent is the whole lesson. A panel radiator sits near 1.3, a finned-tube baseboard near 1.4, an underfloor loop near 1.1. Because $n$ is above 1, output falls faster than the temperature difference does. Halve the water-to-air ΔT and you do not get half the heat — you get $0.5^{1.3} = 0.41$ of it. Scaling linearly is the optimistic mistake, and optimism is how a house ends up cold.

This is why heat-pump retrofits are an emitter problem before they are a plant problem. Drop from an 80 °C boiler to a 45 °C heat pump and the water-to-air difference collapses from roughly 60 K to 25 K; a radiator that was giving 2000 W now gives about 660. The room did not get any easier to heat. The emitter got smaller — not in size, but in every way that matters.