The answer above is a great TL;DR, for the slightly longer explanation, read on:
When building a query, modelling something, thinking about data flow, etc. you'll want to use DataFlow::Node. In general these corresponds to all the parts of the code that can carry a value and therefore make up the nodes in a simple data-flow graph. When using data flow you'll e.g. need to specify sources and sinks as certain DataFlow::Nodes. DataFlow::PathNodes are not themselves directly useable for anything, but rather generated as an output of the data-flow library. When writing a path query the data-flow library will deliver the results in the form of a hasFlowPath(DataFlow::PathNode source, DataFlow::PathNode sink) predicate and a resulting graph that's used to build the paths in the form of import DataFlow::PathGraph. The reason that the library will give this output in terms of a different type is that PathNodes encode more information than just the Node (e.g. call contexts) in order to (among other things) match up call sites and returns properly in the generated PathGraph.