Yep, what they said!
You MAY have interaction between the antennas. Drop one and test again. This is the best test.
---off on a tangent here---
I'd like to model that Jo-Gunn some day. I don't believe they work well at all. An interesting point though, for the "SERIOUS" 11-meter operator, Jo-Gunn recommends the SUPER AUDIO FLAT series which are nothing more than regular flat side yagis with 15KW gamma matches and upgraded mechanical and truss supports. I guess the design of the Smokin' Gun II is for the non-serious operator then. If it was a superior design, they would use it for their serious operators. The hams would use that design too. But, they don't! Why should you?
If you really want to know how well the beam IS working, drop the ground plane and point the beam at a known source. Have the source continue to transmit and spin the beam. Watch you S-meter and note the direction of the beam. When the beam is 180-degrees away from the source see if you have a significant reduction in signal. If not, then your front-to-back is horrible.
This is important, tell everyone you know, an antenna HEARS as well as it TALKS! Antennas don't talk better than they hear or vice-versa! If receive performance goes up, so will transmit performance. If it goes down.....well, you get the picture. I didn't make this up and it wasn't my idea!
Many antenna designs are scaled down to much higher frequencies and tested by receiving a signal while the antenna is rotated. The equipment will then plot out the "pattern" of the antenna. Then they transmit a signal to a receive antenna, rotate the antenna, and plot the transmit pattern. They are practically identical! The higher frequencies (in the GHz range) allows for very small antenna test ranges and makes adjustments bareable. Once they have a consistant design, they scale the antenna back down to the desired frequency and test it again.
Today, almost every antenna is designed in the computer first. Then they make the scaled models. The good antenna companies ALWAYS test in the real world and usually the computer models are pretty damn close (but not always perfect).
You can do these same tests, albeit low tech, and see for yourself! Its FUN too!