Hi Plasma -- I would not describe your task as impossible, but rather not yet sufficiently defined. The way you have set it up, any of those tools would be each poor/good/excellent depending on the problem to which they are applied.
Ease of use, support, etc. are important, but usually subjective considerations, so I won't address that issue here. Two different people could easily have two completely different opinions at the same time and both could be correct.
Analysis error and analysis time are more quantitative and can be evaluated in detail, as I have done for Sonnet in a half dozen or so published papers, mostly in the MTT Transactions. As far as I know, I am the only vendor to have done so.
When doing compartive evaluations (which I do not publish), it is important to make sure they are done on the exact same computer by a user equally skilled in each tool. For example, it is best to use similar meshing. Some tools use edge meshing by default (requires more subsections, but is more accurate, as in Sonnet). In other tools, edge meshing is turned off by default, etc. It is not a good idea to compare an edge meshing turned on result in one tool to an edge meshing turned off result in another tool.
Having done, and helped customers do this kind of work many times, I have concluded that multiple tools are needed for the best results. Even with a narrow set of problems, a well trained designer with at least two tools can do a much better job than a designer with one tool.
After having seen a lot of data from many customers over the 22 years since I founded Sonnet and introduced the first commercially successful high frequency EM tool, my recommendation for the "dream team" of the very best EM tools in each class is:
1) HFSS, frequency domain tetrahedral volume meshing.
2) CST, time domain hexahedral volume meshing (my company represents CST in North America).
3) Momentum, frequency domain unshielded planar surface meshing.
4) Sonnet, frequency domain shielded planar surface meshing (I work for Sonnet).
As just one example of the power of using multiple tools, let's say it is very important that your planar circuit design work the first time. Analyze it in both Momentum and Sonnet. But, let's say, at high frequency, you see a difference. The Momentum (unshielded) result is smooth, but the Sonnet result (shielded) has glitches. Because the Momentum result is smooth, you know that the Sonnet glitches are due to box resonanances. If you put that circuit in a box of that size, your circuit will fail due to box resonances. But you also know that the same thing that is exciting box resonances will also cause substantial radiation if you do not use a box. It will couple strongly to any nearby components, causing a failure of the entire system. A designer who uses two tools knows more about the circuit than a designer who uses one tool.
All the tools have advantages and disadvantages, and that depends strongly on the type of problem being analyzed. It is very important for users to be aware of both strong points and weak points. Disregard, or at least be very careful with, those who discuss only advantages. (Simple unqualified statements like, "XYZ tool is really good," are just plain silly.) I try to discuss both advantages and disadvantages for any tool (including Sonnet) that I recommend. If I can not do both, for any reason, I will not discuss or recommend that tool at all.
So, how many vendors are willing to recommend competing tools???