Problems of hfss simulation and measurement

Dear frieds, I'm designing antenna using HFSS. I want to know how at least many times should the HFSS computed that the simulated results would accord well with the measured results. I usually have iteration number of 7, but with this iteration number, the HFSS usually did not satisfy the delta S of 0.02. To have the measured results accord well with the simulated results, how should i set the parameters in HFSS? If i set the iteration number higher, the computed time would become so long that i could not stand. Could some one give me some advice? Thanks for your kind attetion.

You have answered your question. If solver didn't converge (0.02) I suggest increase the number of iteration.
If your computer can't handle with project, put it here and I'm gonna simulate it for you.

To my opion, the iteration number of 7 is enough for antennas, and in my project, when 7 iterations completed, the converge is litter larger than 0.02, for example, 0.023. Is the converge must be lower than 0.02?
I have a gain of 32dB for an antenna array, but the measured gain is only about 25dB, the S11 is below -10dB, could you think that what reason made the measured gain much lower than the simulated one?

First Of All, there is no specified number of iteration for antenna and anything else......
You should observe the mesh and convergence for particular device. Whats more, if you obtain a value below 0.02 it wouldn't mean a proper result.
In setup options there is possibility to enforce "Minimum converged passes." I recommended to set it on 2. A good habit is to set a "Order of Basis Function" to Mixed order and after final tunning check result witch Second Order Function. Do you use a Descrete or Interpolating Sweep Type?
What type of procces is used to manufacture your antennas?

The antenna is a microstrip antenna array of 256 units. The simulated gain is about 30dB, the measured S11 is below -10dB, but the gain is only 24dB. The difference is so large. What could you think lead the so low measured gain?

You have the array with 256 units and you wouldn't do a simulation with a proper convergance.
Could you show a profile of convergence?

What is the difference between simulated and measuered reflection coefficient?
Did you simulate the array with loss of substrate and conductor?