When you put two signals into a mixer, the IF and LO for example, you don’t just get out the product that you are looking for (the 1×1 or fundamental as it is called). You also get out every other conceivable algebraic product (2 IF x 1 LO, for example). These products are called spurs and a major goal of mixer design is to eliminate them (read all about it in our Mixer Basics Primer). When a customer has a problem with spurs, the number 1 question is: how did you measure them?
Measuring spurs is not straightforward. Similar to measuring Intermodulation Distortion, when you are measuring spurs your measurement equipment itself can mislead you into thinking the mixer is causing spurs, when it is really coming from the equipment itself.
So here is a diagram for how you test spurs and an explanation of all the components: 
IF and RF filters: The most common confusion in spur testing comes from not filtering the output of the synthesizers. Any tunable synthesizer will produce some amount of spurious content (in addition to producing fIF for example, it will also produce f2xIF and f3xIF) and if these higher order terms are not filtered out they will significantly affect the measured spur levels.
Spectrum Analyzer Attenuator: Every spectrum analyzer has a mixer at the front end, and this mixer causes intermodulation distortion (IIP3 as well). This distortion is reduced by placing the attenuator in front of it. You might ask (I did), ‘why not just turn down the IF power into the mixer down to get the output power down?’ The answer is that spurious products don’t increase or decrease linearly with an increase or decrease in power the way the fundamental does. Higher order IF spurs (i.e. 2 IF x 1 LO,3 IF x 2 LO, but not 1 IF x 3 LO) will increase with power according to (n-1) times the increase in input power in dBs. So if the input power increases by 3 dB, the 2 IF x 1 LO will increase by (2-1)*3 = 3 dB, while the 3 IF x 4 LO will increase by (3-1)*3 = 6 dB. This is why it is best to put maximum power into the mixer and then attenuate the output – because it makes it easy to see the spurs.
IF Input Attenuator: This is the least important element. The mixer will act as both a source and a sink for IMD products, and some of these will come out of the IF port, reflect from the filter, and return to the IF port to cause further distortion. This attenuator will reduce this affect.
How important are each of these elements? There’s an easy way to test. Try testing without a given element, then add it in and adjust your powers so you’re at the same place. If you see the measurement change, then you know that it made a difference.
Happy testing!
Good write-up on the basics. I might add that a one good way to help determine if observed spurs are generated in the mixer or in the spectrum analyzer is to swap the 10dB attenuator for a 20dB attenuator. If the observed spurs drop 10db like the fundamental does, then the spurs were likely from the mixer. If they drop by 20 or 30dB or more, then they are likely generated in the spectrum analyzer.