That may be, but we still need to define the problem in order to investigate the problem. You'd be just kicking the tires around otherwise. It could well be a reiterative process of investigating - learning - redefining as you said, till we get the full picture. And starting the gradient walk from a conservative point is usually a better strategy. Unless you are reasonably sure that two different manifestations are the same phenomenon. We know, for, example, severe and mild cases of ME/CFS are two different presentations of the same because patients go from severe to mild and vice versa all the time. So it would be prudent to include both severe and mild cases in investigation for more complete picture. PEM with or without delay, on the other hand, we are not so sure at this point.
It's different clinically, obviously. The aim there is to provide treatment and reduce the suffering rather than figuring out what it is. So, it would be better to be liberal than conservative there.