OK, here we go again.
If you test 20 samples, and 19 fail... you should be asking "is there anything wrong with my sample selection?". "Did I actually choose 19 samples of the same malware family?" etc.
There are hundreds of thousands of samples out there that the particular product doesn't detect, no matter what product it is... so unless you're pretty damn sure that your selection is random, well representing the overall situation (or not - sure, you can be interested in your local malware, in which case the sample set should represent that, of course) - then no, the result is meaningless, and the correct reaction is "so what".