I can see why you came to the answer $(-\infty,-11) \cup (-11, \infty)$. That's because if you compute your $f\circ g$ as a symbolic expression, you get: \[\frac{1}{\frac{x-7}{x+5}-3} = \frac{1}{\frac{-2x-22}{x+5}} = \frac{x+5}{-2(x+11)}.\] On this face of it this is perfectly well-defined at $-5$.
However, if you decide that $g$ is not defined at $-5$, you might then conclude that $f\circ g$ (the function "apply $g$ and then $f$") can't be defined at $-5$, since after all you can't apply $g$ in that case.
A relatively sophisticated perspective on this is that "$f \circ g$ has a removable singularity at $-5$", that is to say, a point at which the function is technically not defined, but for which there is a perfectly good value (and indeed, only one value) the function could have that would be consistent with its behaviour elsewhere – in this case, you can extend $h$ by setting $h(-5) = 0$.
In fact, we tend to come across this kind of situation all the time, and usually we just plain ignore it – e.g. when cancelling $z$ from $xz/yz$, we rarely bother to say "except when $z = 0$". It's rarely problematic to just remove all removable singularities as soon as they turn up: having a function defined on a bigger set than you strictly need isn't a problem because you can always just restrict it to a smaller one.
However, in a question like this which is specifically asking about domains, it pays to be precise – and that means that whenever you replace a function with its definition, you must ensure that the argument belongs to its domain, possibly restricting your own domain as necessary.