For those interested, a while back I did an experiment that was related to a part of this discussion. I may have to do some digging if anyone has questions, but I will post the details/results of the experiment here.
I started with two 5/8 wavelength antenna models (not an Imax model, but a bit shorter). These models consisted of an upper 5/8 wavelength section, a lower mast section that went all the way to the earth below, and one of them had radials while the other did not. I modeled the antennas at 128 different heights between just above ground level up to 11 meters, all of the heights were spread equally apart (by about 3.4 inches). This was also over moderate ground... I recorded various data into an Excel document to plot the differences over the varying heights.
I'll let you guys draw your own conclusion from said data...
One thing that struck me when I did this was how stable one set of models was over the change in heights, compared to how volatile the other set of models were. A set of horizontal radials really does stabilize many aspects of a 5/8 wavelength antenna system...
Anyway, lets start with gain, because everyone and their brother loves talking about gain...
And, with gain people will inevitably talk about "Take Off Angles", so lets do that next.
Its worth noting here that for the antenna with no radials, two points here a higher angle lobe formed, and then became dominant. The first time it happened, the existing lower angle lobe got smaller and smaller until it disappeared, and the higher angled lobe got lower and lower and eventually matched the angle of radiation that the radials model had. Then another high angle lobe formed, took dominance again, although the process didn't actually repeat itself. In this case, the higher angled lobe got lower with height, and actually came close to merging with the existing lower angle lobe. As they got close together, the low angle lobe regained dominance.
To round up the performance part of these lets do radiation efficiency next...
Onto tuning data... Lets start with the R variable over the range of said frequencies...
And X comes next...
And finally SWR (before any form of matching would be applied to said models).
If you love analyzing things like I do, eat your heart out...
The DB