When determining the structure of a material, one should not assume that when a good, or even excellent, fit to the experimental data is found that this indicates that the structural model is correct. There may be other models that fit better that have not yet been found. Even if it is known that the model offers the best fit to a set of experimental data, it may be that with more precise data, it would be clear that a different model is to be preferred. There is also the problem of “fitting noise”: Sometimes, in order to improve the fit, refinement will distort a model in ways that are not chemically reasonable. If refinement improves the fit to data by distorting a phenyl group so it is no longer planar and the C-C bond lengths are unequal, that model would be far less plausible, even though it provides a better fit to the data than a chemically reasonable model.
Perhaps a better question than “what is the correct model?” would be “what is the most accurate structure that is consistent with a set of data?” since with better data, one may be able to learn more. A key question will be to establish the best symmetry description for the structure. Crystallographers have traditionally been very concerned about missing symmetry. When a structure is refined in too low a space group, the missing symmetry will invariably degrade the quality of the refinement. Historically, the highly respected Caltech crystallographer Richard Marsh, who died in 1994, published reports on crystal structures where symmetry had been missed. Invariably, the higher-symmetry structure was preferred, and often removed implausible aspects in chemistry. Dick was a pleasure to be around. I am not the only one who misses him. The process of finding a higher symmetry space group for published work is often referred to as “Marshing” in his honor.
If the space group used for the model is too high, then the structure will be inaccurate because the symmetry will force details of the structure conform to symmetry elements that are not present, for example by requiring elements to lie on or be reflected by a mirror plane that is not actually present. When the symmetry is too high, the model will miss key details in how the structure deviates from that falsely imposed symmetry. When it is found that a structure is in too high a space group, there are at least a few of us that call this “Anti-Marshing”. While data quality limits the ability to determine structures, my personal feeling is that using too high symmetry is actually a bigger problem than missing symmetry. As an example, reporting that a material is centrosymmetric when it is actually acentric will cause some interesting science to be missed.