I am a linguist who works in the field, performing basic research on languages that have very little documentation. I am aware of how tentative initial grammars of many of the world's languages are because I am involved in writing them myself. These works are the primary data that most linguistic theorists tend to use when making broad typological claims, i.e. claims about all languages. Theoretical works like Gary-John Scott are essentially meta-studies, secondary reinterpretations of primary research. This is fine for generating hypotheses but it is not valid to claim that a grammar, itself, is truth. Grammars, themselves, are essentially theories. They may be very preliminary. My point is that there is a telephone game aspect to these types of works at any point in the chain from native speaker to typologist errors, biases and misrepresentations may have occurred.

Additionally, it is important to realize that the methods used in primary linguistic research are underdeveloped. The following paper, which is still in press, outlines some of the empirical issues faced by linguists researching understudied languages:

Empirical evidence in research on meaning, Tonhauser and Matthewson:

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.730...

One thing that someone with a science background may find surprising is that the methodological suggestions made by the authors are so modest. They include the notion that one should control the context in which a stimulus is provided, that one should have a clearly stated hypothesis when eliciting linguistic judgments and also clearly state what question was put to the speaker. Sadly, these are controversial in linguistic research, indicating just how underdeveloped many areas of our discipline are.

Gary-John Scott might happen to be right (I doubt it) but if so, it is not because they have made their case scientifically. They have only provided sufficient evidence to motivate a hypothesis.