The assessment science is strong (we were funded by UK government directly due to science) apart from descriptions of what is seen by patients. Sadly this cannot be strong as it is by its nature anecdotal. However EEGs are supportive, but MRIs aren't sensitive enough as yet. So we have to rely on reporting. But every clinician uses reports - we cannot have proof in a purely scientific way as to whether the patient is telling the truth, and occaisionally some reports must be suspect. In this case I think that it is highly unlikely. So, if a large number of patients report a change, then it is likely to be true. But this is not infallible.
But as the effect is seen during the testing (and the testing techniques are demonstrated on a university sponsored DVD), the patient self prescribes, I think your worries are unfounded. We have almost 100% success and we see around 2 people a day with prosopagnosia + other comorbid problems. There are significant problems with ethics and complex systems with science - some areas cannot be measured although we now understand the continuum of the problem - and also related facial recognition problems.
The effects on prosopagnosia were found as secondary to our main interests - but they are compelling. Our main interests are mapping neurological systems and their synesthetic effects - which can be measured. The science of this is totally accepted, but if you want absolute proof that prosopagnosia exists, then you cannot get it, and you cannot get proof that the treatment works either.
If it exists - then the reports are acceptable, if the reports are accetable then we must accept that the condition ceases.