I ran a few models with the 4nec2 modeling software to see the radiation patterns for two end fed half wavelength antennas 5 meters high over "real ground" at various distances apart. They are also in phase. This is what I found...
1/2 wavelength apart...
5/8 wavelength apart...
3/4 wavelength apart...
7/8 wavelength apart...
1 wavelength apart...
At a half wavelength apart there are two lobes broadside to the pair of antennas. As you begin to spread the antennas apart from there the existing lobe begins to shrink as new lobes open in line with said antennas (on the X axis). These lobes continue to get larger, and if continued past the point shown here begins to split into two lobes each.
Because of this the distance of separation matters. If you are looking for two deep nulls 1/2 wavelength separation is your best bet. If you are looking for maximum gain in two directions 5/8 wavelength apart is your best bet, although a 3/4 wavelength separation will yield 0.06 dB less gain, this is insignificant and will not be noticed by anyone.
These look just like the diagrams shown in the last few versions of the ARRL Antenna Book for antennas with similar spacing apart. It was listed in said book to be for quarter wavelength antennas and these models are of half wavelength antennas. I guess size does not make much of a difference here.
Well there was my first modeling project... I learned a thing or two in the process and hope it is of help.
The DB