Well, my friend, that's where the antenna engineering takes place ;-)
As you probably know: Gain = Directivity - Loss, and if you check the directivity, everything is okay (DirRHCP > 7dBi), thus you have about 10dB loss (just an assumption, you can see the actual value by plotting the RadiationEfficiency (under the AntenneParameters in the FarField-Plots).
Most (practically all) of the loss is due to the dielectric imperfection (finite isolation -> dielectric loss). There are only two ways to reduce dielectric loss:
- use a better dielectric (but your Rogers is already pretty good)
- minimize the electric field areas and intensities.
The first you can check by right-clicking on the Rogers Ultralam material, choosing "Properties" and then cloning the material and setting the loss tangent to zero and simulating the whole model again.
The second is more tricky (but the way to go here, I think). To minimize the field intensity, the distances between the arms and between the arms and GND would have to increase -- which is probably not too easy to do. On the other hand: why are your arms so long? I redid your simulation with n=1 (just one turn) and got a good result (5dBi) right away (of course there is a trade-off here, which is probably the polarization ratio, here, I think... but I didn't investigate that any further).