If you post your files compressed with zip, gzip, bz2, p7z or any other compression algorithm which is public, I will look at your files. But I refuse to open rar files unless I really have to, since the compression algorithm is not public. (It's just one of my policies!)
When the antennas are close, you will get coupling between them, so their characteristics will change. I would attempt to simulate it as as two individual antennas, ignoring any power splitter you may have. Then add that later.
I believe the sepparation can be smaller than the free space value due to the dielectric, but I must admit, I am not sure about this. I thought that was so, but then my next point about capture area makes me wonder if this is true.
I don't believe the spacing needs to be less than lambda/2. If it is too close, you wont get any gain. You want the capture areas of the two antennas to just touch. Theoretically, at least in free space, the array factor will tell you the gain. But in practice, if the antennas are close, their presence will effect each other, whereas the array factor is derried by assuming the antennas are isotropic radiators.
Sorry I can't be more help. I'm no expert, but I try to help. If I'm wrong on anything, I'm sure someone will correct me, and we will both learn.
Deborah