RFC-4689 defines how to measure jitter. However, there is no specification to define what is nominal or acceptable jitter. It is application dependent.
However, you can use the long term max jitter as the design criteria to determine whether a particular application will work in the network.
Moreover, jitter is a measureable network characteristic. However, it can vary greatly under different condition. In general, physical links do not cause jitter. Router/switch/firewall processing cause jitter. Link aggregation and link conversion cause jitter. Packet inspection and packet encapsulation/encryption cause jitter.
The best way to characterize jitter is by running long term end-to-end test and use the max jitter as the design criteria. You mention that you only have visibility to sections of the end-to-end connection. You can use the statistical Root Mean Square to combine the jitters at different segments to estimate the known jitter. However, you still have the unknown segment to deal with.