What used to be called 'Reference Distribution' is now in 'User Data' in the FEFLOW GUI. So a 'nodal reference distribution' is a nodal User Data distribution that you can set up with a user-defined name (you need to use 'VaryingSaltwaterHead' in this case). Such User Data distributions are often used for plug-ins as they are an easy way of defining data in FEFLOW not belonging to the standard set of parameters. In the plug-in, the identification of the distribution is done by name, which is why you need to use the name mentioned in the instructions.
Please note that the plug-ins that are not part of the FEFLOW distribution come without any warranty. They typically have been created on basis of a project need in our own consulting work or by support staff quickly responding to an urgend client need. So they are usually not as thoroughly and formally tested as the software itself is. Regarding a model example: I'm sure I had done a simple one at the time I coded the plug-in in order to ensure the applicability, but I haven't kept it. Thus I recommend you do your own little box model in order to understand the plug-in functionality.