Check the FCC reports from your test house.  They will probably list the equipment that they use (preamp and antennas are the main items) for testing.  You'll probably be shocked at how much they cost new, but you might find them used for a reasonable price.  

You can do a lot with a calibrated broadband antenna.  A log-periodic can be bought that goes from 20 Mhz to to 3 GHz, for instance.  You go outside of your building, power up your dut, and use the antenna at the prescibed distance and at both polarizations.  You just hook it up to the spectrum analyzer.  

I have NEVER found an antenna used on ebay that has its Calibration antenna factors!  Good luck trying.  You CAN find antennas on ebay that are uncalibrated, but they are pretty much worthless for any EMC precompliance testing.  They are usually old and beat up/missing parts and manuals.  A few dB difference in received power makes all the difference in the world, so it is probably worth it to get a new antenna, without any antenna element corrosion or damage, and a recent antenna factor chart.  Otherwise you might be fooling yourself and wasting a $3000 visit to the test lab.

You can get calibrated broadband antennas from places like this:

http://www.ahsystems.com/

I'm not familiar with the MIl std you refer to. I am used to the european CE requirements. To meet the Class B limits you have no chance reading the values with a spectrum analyser and antenna. The background noise levels will be much higher (30dB) than the signals you are looking for. This is why open field sites are in the middle of no where. I had to get my own shielded chamber constructed. I also had to oput a 30dB broadband amplifier on the front of the spectrum analyser to get enough sensitivity. If you use the amplifier in the open it will saturate on the ambient signals unless you use an RF pre-selector. You also get more sensitivity by running the equipment stupidly close to the antenna but it is still not enough! If you can see your equipment emissions in the open easily, it must be absolute rubbish on emissions.

This relates to freee field measurements. If you are doing "conducted emissions then you just need a LISN and a quasi-peak detector on your spectrum analyser.

That's probably MIL-STD-461 or 464

TTFN