I believe patients are scared that these patients will 'pollute' the cohort. I very much disagree and I believe that we will find that with emergence of precise testing and biomarkers such as 2days CPET, viral markers at disease onset, tilt table test, there will be patients who suffer from both mental and physical illness.
That said, it is absolutely necessary to have clean cohorts in the early stages of research. Confounding factors must be eliminated in order to get at the fundamentals of the condition. Once the basics of the condition are understood, it is possible to include patients with confounding factors to see how those confounding factors affect symptoms, progression, treatment, and so on. No, the earliest research does not fully describe the condition in all it's ugly glory. It never does. Look at the early HIV research. It (unintentionally) did not include the majority of HIV-positive patients. The purpose of the earliest research is to hunt down the core features, which are most likely to be found found in the most severe and unconfounded patients.
Eliminating confounding factors is not prejudice, it's sound research -- as is carefully exploring the impact of confounding factors once you know more about the condition.