I often have get men on the Peak Testosterone Forum trying to figure out why their testosterone tests are varying so much. Another fairly common issue are clearly impossible lab results that might just as well be tossed out the window. These incidents bring up the whole issue of accuracy and just how much stock one should put into any blood draw.
Before I go on, I want to say that I am a HUGE fan of testing, including self-testing. Prices have come down dramatically and doctors rarely pull enough numbers due to cost constraints and insurance limitations. So we often need to pull our own numbers. But, by the same token, one also has to be realistic about accuracy levels: even though these lab tests use very sophisticated and expensive equipment, they definitely seem to have their limitations. Below are Four Examples Where Testosterone Tests May Not be Very Accurate:
1. Total Testosterone Test Variability. Here in the U.S., our testosterone results are typically given to us in 3 significant digits, i..e 841 ng/dl or 337 ng/dl, etc. However, this is highly misleading. Check out what one recent urological study found: 
"Despite advances in technology, inconsistent laboratory practices, among other issues, leads to unreliable blood test results. "In some cases, testosterone (T) levels, tested on the same day from a blood sample taken from a single patient, differed by as much 30 percent from one lab to the next," says the study's lead author, Dr. Paduch, a urologist and male sexual medicine specialist at NewYork-Presbyterian/Weill Cornell Medical Center and associate professor of urology and reproductive medicine at Weill Cornell Medical College." 
I recently pulled my total testosterone with one hour of each other from two different labs and one read was very close to 700 ng/dl and the other 800. This is a 15% difference. Of course, some differences should be allowed between labs and the usage of different equipment and methodologies. But these kind of differences do seem disturbing and this is the reason that the authors above were pleading for industry standardization.
2. Topicals / Transdermals on the Forearm. For years I was on a compounded testosterone cream and my testosterone reads were always in the 480 - 520 ng/dl range. However, one time I had a crazy-high pull of 1400 ng/dl or something close to that. It clearly was an outlier. One of our Peak Testosterone Forum members had something similar, where his total testosterone more than doubled from 400 to about 1000. He wrote:
"This low e2 with high free and total T confused me and Peak too. I realized the mistake I made, I used my forearms to apply the gel, then had a draw after. Yes I washed my arm, but this isn't enough as it is still below the skin level where the needle went. So my readings on gels were all false due to this." 
In this case (and mine), the poor read is not actually the laboratory's fault but rather a miniscule amount of topical testosterone left on his forearm that was drawn into the needle as it went through the skin for the blood draw. Men who are applying testosterone to their forearms, as I and the above forum poster were, do have to be careful on testing day to insure accuracy.
3. Bogus Free Testosterone Exams. Men have very low free testosterone comparatively, because free testosterone is generally only about 2% of total testosterone. Although LC-MS/MS technology is used, which should be accurate enough to handle male free T readings, I see a very common issue on the forum: ridiculously low values. Because of these issues, I encourage men to use total testosterone and SHBG and just calculate it that way on this pagewith an online calculator. For details, see this page: How to Measure Free Testosterone. One of our posters wrote the following which made me feel better:
Did you know you can inexpensively do your own testing for most hormones? The industry leader is Discounted Labs..
"Dr. Shippen still doesn't trust the free T tests. He still uses total T and SHBG to calculate free T. Maybe his view is justified." 
Dr. Shippen is one of the most famous TRT physicians and has a classic book out about the subject: The Testosterone Syndrome
By the way, I am NOT saying the test is always inaccurate. I have seen the issue repeatedly though. I would also add that these outlier free testosterone values are typically very low and simply could not be achieved assuming any reasonable level of total testosterone, SHBG or albumin, the three variables that essentially determine free testosterone.
4. Dirunal Variation in Testosterone. Men not on TRT will have their highest testosterone levels in the morning. Some physicians forget to tell their patients this and their doctor, and the unsuspecting man will get his testosterone drawn in the mid or late afternoon. As I explain in my page on Daily Testosterone Levels, T can fall by about 35% for a healthy young male from morning to evening, so this can yield inaccurate results.
CONCLUSION: Pulling testosterone is VERY important. However, it should not be considered a highly accurate test, and I think one should not be surprised if your testosterone varies 10-20% just due to "inaccuracies." However, this is not as important as you might think for the following reasons:
a) Symptoms are just as important as the lab reads. If you feel good, and your lab says it is at 600 versus 700, does it really matter? Or, likewise, if your morning erections are restored with a lab read of 600 versus 700, does it really matter?
b) It is more important to go by a series of blood draws in my opinion. Testing has now because so inexpensive that there is no good reason not to pull your testosterone quite regularly. See my on Inexpensive Testosterone Labs for more information.
c) If you are very high or very low with testosterone, it is important to know. The reason is that side effect risk, some potentially risky, increase significantly at the extremes..
2) Urology, May 2014, 83(5):980 988, "The Laboratory Diagnosis of Testosterone Deficiency"