Hi,
I made a Webench-Simulation, the ICThetaJA Effective is given with 6°C/W (Effective IC Junction-to-Ambient Thermal Resistance).
In the datasheet the value RθJA Junction-to-ambient thermal resistance is given with 42.9°C/W. I dont think that this value can be reduced such
like this, also with a perfect layout design (I dont know what kind of layout Webench use for this desing). I hope according JESD 51-7.
br
The theta jA effective parameter would be done per the power loss of the test.
A more appreciable curve to use (in my opinion), would be figure 8-7 and figure 8-8 in LMR33620-Q1 D/S. This is based on experimental data. This gives you theta jA per available copper area and max ambient your system can support at full output power.
It looks like theta jA saturates around 45degC/W, which is a little surprising to me per my earlier assessement....this die may not have been designed as well as other dies I am used to working with that utilized VQFN package technology.
If your application is stressed for thermals and you can allow for a little bigger solution size with this device I would recommend the leaded orderable in the LMR33620 D/S (no Q1).
Alternatively, the newest generation of this device has even better thermals and comes in WSON for best of both worlds for small size and thermals...LM63625/LM63635.
Let me know what you think. Hopefully I did not make this more confusing!!! 
