The best available data often are buried in the studies done by test manufacturers prior to FDA approval and don't necessarily see the light of day in scholoarly publications. But even these data are not precise. Scientifically rigorous research to determine exact window periods is essentially impossible from practical, ethical, and cost perspectives. 

[Therefore, our] window period advice is based primiarly on biological principles.... We have good data on time after exposure when HIV antigen or RNA are detectable, and when measurable antibodies appear. On this basis, the 4th generation tests are virtually always positive within 4 weeks and remain positive thereafter for the life of the infected person. It is also the case that there are few if any published case reports of people with negative test results at 4+ weeks who were later found to have HIV; and all the HIV/AIDS experts we know also say they have never seen this in any of their patients. The same sorts of considerations apply to combinations of tests, such as when an RNA test is done at say 2 weeks and a standalone antibody test is done at 4 weeks.

On the other hand, some experts and agencies are uncomfortable standing only on these biological principles and hold out for more definitive data, without which they take consiervative stances. Often there also is a tendency to not change previous guidelines unless and until new data conclusively prove that information was wrong, which explains why 3 months remains a commonly accepted standard, based on older standalone antibody tests. And health departments and other governmental agencies also tend to take conservative stances, preferring to err on the side of longer advice about window periods. CDC is in this category.