Model vs Reality

I took some measurements of my droopy 40M vertical today using my MFJ antenna analyzer, and the results are nowhere close to what the EZNEC models predict. Here are measurements taken at the antenna itself (through a 2 foot patch cable):

Frequency	Rs	Xs	SWR
7.0	142	165	5.8
7.1	142	173	6.0
7.2	142	181	6.3
7.3	140	189	6.6

The EZNEC model I made for this antenna shows that it should have an almost purely resistive load of about 50 ohms in the middle of the 40 meter band. So why is the model so different from the measured values? Obviously, my model is wrong. Some things I can think of:

• The model doesn't include the tree that's supporting the antenna.
• The radials in the model are straight; the actual radials are not.
• I tried removing some of the radials from the model (from 6 to 3), and this shifted the resonance point, but that's all.
  
• The radials in the model are uniform length; the actual radials are not.
• I tried randomizing the radial lengths in the model; again, this shifted resonance, but that's all.
• The antenna is operating as an elevated vertical, and therefore the radials need to be resonant (variation on the previous theme). Idea: unhook the existing radials and attach two radials cut to 1/4 wavelength; observe result.
  
• The model has the radials elevated 0.1 ft, when in fact they are buried about an inch deep.
  
• My ground model is not correct.
• There's some physical fault in the box I constructed (it has an SO239 on one end, and two bolt/washer/wingnut combinations I use to attach the coax to the wires).
  

At the other end of the coax (100 ft of RG-8X), the readings are:

Frequency	Rs	Xs	SWR
7.0	54	54	2.5
7.1	43	56	2.7
7.2	32	55	3.0
7.3	24	50	3.3

(goes looking for Transmission Line chapter from ARRL Handbook...)