First of all, most "universals" aren't universal. They're tendencies, statistical correlations. See e.g. the at the University of Konstanz (the archive itself seems to be down again, but bookmark that page and check back later) or by Matthew Dryer.
As for those universals to which no exception is known (such as, presumably, this one about stops), well, languages violating them will probably always be rare, but there is only a finite number of languages currently spoken on Earth and not all of them have been well described by linguists. Until the late 1970s there were no languages known in which object preceded supject and verb in basic word order. Then Hixkaryana (OVS) was discovered. There are only a very few OVS and OSV languages reported (and in several cases the basic order is disputed), and none of them have very large populations. It's easy to imagine a world in which these languages died out without being studied, and then we would not know that they were possible. But thanks to Hixkaryana we do.
It's probably a good idea to know which features of your conlang are typologically unusual, and to remark on it in your grammar. But you shouldn't necessarily think of them as unbreakable laws. Rather, they're probabilistic, and a feature can have a finite probability which is sufficiently low that it isn't guaranteed to show up in a sample of 6000. (Of course, there may be some truly absolute universals too. But you can't identify them simply by observing that a feature doesn't exist in any documented language.)