You can not use AGT over real earth, but you can use perfect ground. When running AGT it needs to use either free space or perfect ground, anything else will not be lossless, and therefore will give a bad AGT result. 4nec2 does this check for me, if there is a ground present, any ground, 4nec2 will substitute a perfect ground. I can't speak to EZnec on this, thats your department.
Bob85 isn't the only person who uses dBm when it comes to measuring antennas. For nearly a year now my preferred way of tuning antennas uses dBm readouts from a portable spectrum analyzer. Its no different than how I used to use field strength, only now I have specific measurements rather than a relative number that I could adjust with a variable resistor. The setup is nothing fancy, its just a portable antenna mounted far enough out with a coax running to a spectrum analyzer that is plugged into my laptop. This one requires a laptop as my spectrum analyzer has no built in screen. I am effectively tuning the antennas by measuring performance directly.
When it comes to dBm and the 4nec2 modeling software, the results are in fact not accurate. The reason is how the angles form in models over an earth. Most people assume those angles form immediately at the antenna, or immediately after the near field collapses, but unless the antenna is mounted at earth level or immediately above said earth, this is an incorrect assumption. Unfortunately the 4Nec2 modeling software makes the same assumption. However, I can't speak to the EZnec pro software that Henry used, so I cannot say for sure weather or not his data is accurate or not.
Henry's chart shows things differently than 4nec2 shows them Essentially, the numbers across the top are the height of the measurement in meters, while the numbers going up and down are the measurements of signal strength in dBm, and as these numbers are negative, the lower numbers are better. From two sentences above the graph it is stated that these measurements are from 30 Kilometers out from the antenna. I think the layout of this chart could have been made more intuitive than it is here.
I have not modeled a vector with horizontal radials in addition to the cone. In my view, the currents flowing the basket area of this antenna are necessary and required for this antenna to function as intended. Adding horizontal radials will cause currents to be divided between the cone and said horizontal radials, this will change (lessen) the currents on the cone, which will affect how the cone does the job it needs to do effectively (in your case I would say that this effects the cone's ability to cancel the the out of phase RF from the antenna within the cone). Is this desirable? To be honest, I haven't explored this idea, but conceptually to some extent you should be able to get away with it removing currents from the cone of this antenna, up to a point. I am not sure if adding four horizontal radials will get you to that point or not, but I am pretty sure they will get you awfully close. Mind you, this is mostly just ramblings of ideas from off the top of my head and not completely thought out, so at best the ideas in this paragraph should be explored more thoroughly...
Long story short, when to see a dB or decibel figure, a decibel is always in comparison to something else. dBi is decibels compared to a theoretical isotropic radiator in freespace, dBd is decibels compared to a center fed half wavelength dipole mounted in freespace, dBq is decibels compared to a 1/4 wavelength ground plane antenna with four horizontal radials in freespace, dBm is decibels compared to 1 milliwatt of power. You can't have just 5 dB gain, or the 9.9 dB gain as the a99 claims. A decibel always have to have a point of reference or comparison. As there is no assumed reference when it comes to decibels, if you do not specifically state a reference or comparison like I did with the examples above, or dB gain over x antenna, or something like that then the number you give in meaningless, and is in fact not technically a decibel, even if you say the word.
The DB