The definition you provide actually says otherwise:
"A theory is a logically self-consistent model or framework for describing the behavior of a related set of natural or social phenomena. It originates from or is supported by experimental evidence (see scientific method). In this sense, a theory is a systematic and formalized expression of all previous observations, and is predictive, logical, and testable. In principle, scientific theories are always tentative, and subject to corrections, inclusion in a yet wider theory, or succession. Commonly, many more specific hypotheses may be logically bound together by just one or two theories. As a rule for use of the term, theories tend to deal with much broader sets of universals than do hypotheses, which ordinarily deal with much more specific sets of phenomena or specific applications of a theory."
Theories are constructed around facts/observations, but they are not constituted by them. They are abstract, theoretical frameworks which aim to explain and make sense of facts and observations. For example, we haven't observed all of the theoretical objects posited by quantum field theory (at least, not in the way in which we might observe a falling apple) but the theory can none the less be used to explain observations and make testable predictions.
Also, the contents of theories commonly aren't empirically verifiable (though, importantly, they should be falsifiable). In particular, most theories make universal generalisations - i.e. with the very rough form of for every object x, satisfying some set of conditions Y, some description/outcome z will apply/occur. Absolute empirical verification of this involves observing every such instance and confirming that the rule holds. However, we are but limited beings (spatiotemporally) and so we have to make do with more approximate methods. Thus we form theories which are consistent with our observations and attempt to find further confirmatory evidence and, more importantly I feel, we also test whether they can be falsified.