It scales burning rate and time. At ignition scaled time is 0 and the burning rate is zero. The FDS timestep is scaled by the ratio of the current heat flux to the cone heat flux. So if the FDS timestep is 0.01 s, the cone flux is 25 kW/m2 and the FDS flux is 50 kW/m2, then the scaled timestep will be 0.02 s. FDS then looks at the cone curve at the scaled time and multiplies the burning rate by the ratio of the current flux to the cone flux. If your flux was always twice the cone flux, then the fire will be twice as big but burn half as long.
As for how well it works you can get a feel for this with Figure 11.1. Burning behavior of materials isn't truly linear like this very simple model assumes. Going from 50 to 25 or from 50 to 75, the FDS predictions are not perfect matches to the cone data at those fluxes. The 75 is pretty good overall but the 25 doesn't capture the delayed ignition at the lower flux.
There is no way to predict how well this simple model will work for for your specific case. In general, would expect if your FDS predicted fluxes over time stays close to the cone flux that it will be reasonable.