A metatheory is a theory about other theories. Those other theories and their constituent elements are the "data" on which metatheorising is based. So, in building metatheory we need to draw a boundary around the kinds of data (other theories) we are interested in exploring. This boundary defines the domain of the metatheory. It doesn't matter how big or small that domain is, as long as we draw it and clearly describe it. Without any boundary around the range of relevance of the metatheory it cannot be tested and it cannot be validly argued that it accurately represents its data. No domain boundary equals no science.
The size of the domain doesn't matter in the building of metatheories. They can be really big, e.g. a metatheory for human development, or relatively small, a metatheory for behavioural approaches to treating phobias or a metatheory of green building design:
http://www.metaefficient.com/miscellaneous/unified-metaefficiency-theory
This point gets lost on lots of people who discuss metatheory. See this post for example:
http://dumbscientist.com/archives/theories-and-metatheories
The author of this blog decides what is a metatheory based on how big its domain and almost conflates that with the spatial reference of the theory. This is a very common misunderstanding of the nature of metatheorising. Under this misconception Big Bang theory and Darwinian evolutionary theory count as metatheories because they are so big in their domain of relevance but they are not metatheories. They do not meet the central definition of metatheories - They are not theories of other theories. They are theories of the empirical world of physical matter and biological speciation respectively. The size, i.e. the scope or domain, of a theory is not a criterion for acceptance as a metatheory.
Size of domain does not matter when figuring out whether a conceptual framework is a metatheory or not.
Where size does matter is when metatheory is developed that is specifically intended as applicable to a large domain of disciplines and fields of scientific study. This is usually the case with what have been called integral theories and metatheories. I use the term "integral" in reference to the long tradition of meta-level thinkers and researchers (Roger Bacon, Vladimir Solovyov, Pitirim Sorokin, Jean Gebster, E. F. Schumacher, Ken Wilber, Bill Torbert, Ervin Laszlo and many others) who have tried to develop overarching big pictures through the accommodation of many other theories and systems of thought. Integral metatheories have very ambitious domains of relevance - all theories of personality, all theories of change, all theories of human development, all social science theories, all theories of spirituality, etc.
The problem with integral metatheories is that they often lose sight of the necessity for drawing any domain boundaries at all. They become a little too ambitious in the scope of relevance for their frameworks. They generalise their ideas way beyond the domain boundaries that define their data set (if they ever collected any data, i.e. sampled other theories/metatheories).
So size of domain does matter when we try to build integral metatheories. However, when we set out to build these huge integrative frameworks we need to be extra attentive to the issues of domain and boundary setting. We had better make sure i) we have clearly identified the domain of the metatheory, ii) we have sampled data across the whole range of the domain, iii) provided a rationale for our sampling process, and iv)know when we are making solid claim about the generalisability of the metatheory and when we are making more speculative meta-conjectures.
As far as I can tell Wilber's AQAL metatheory falls short on all these counts. For example, how can an AQAL-informed researcher respond to the question - "How do you know that AQAL accommodates all the major theories of human development?" The only answer that they can give to this point is that "Wilber says so". Any metatheoretical system that relies on this kind "argument-from-authority" response is not scientific. (See Wikipedia's page on this:
http://en.wikipedia.org/wiki/Argument_from_authority
So domain size doesn't matter when we want to define a metatheory and it does matter when we want to define an integral metatheory. But whatever the type of metatheory we are researching, we had better pay special attention to defining its domain if we want our metatheories to be based on scientific evidence as well as idiosyncratic insight.