In the usual place where losses go?

Yeah but not quite. The way to use the atmospheric loss factor is first you have to have calculate radar range in vacuum, then use the result to calculate the atmospheric loss factor, Then you use that resulted loss factor to calculate the final range estimate. The graphics alone wont be enough.

But anyway, i guess we can make simplification as one in other George M Siours's book "Radar Performance Modelling 2nd Edition"

Since our interest is the loss factor and we have radar reference range of say 400 Km. The equation would look like this

R=Rref*(1/Llens*LWeather)

From the graph you provided the loss for 400 Km range and some 2 Degrees of Elevation is about :

LLens : 0.7 dB

LAtmos : 2.8 dB

So the range with those factors are

R=400*((1/(1.17*1.90))^(1/4))

R=400*0.82

R=327 Km.

So the range "loss" is about 18% for the elevation angle specified. One have to question however at what elevation angle the radar was tested.

OR maybe if Irbis were to put on the ground and use same parameters as above, considering the loss was only 18% in terms of range that still not really answer the question of near

50% differences in range. Or does the S-400 radar simply tested 0 degrees and "look ahead" to the Horizon which dont seems to be practical. But from the graphics does provide over 33% range loss. But then the range against 3 Sqm would be about 267 Km, not 232 Km calculated from the Export brochure.