The only idea would be that there is a structural mismatch between the SDF and the netlist you had read-into PT.
The message indicates that the netlist does not have a direct connection between the two cell-pins it mentions.
Have you tried to ...
-> Pull the netlist into an editor ...
-> Pull the SDF into an editor ...
-> And then *manually* cross-reference the offending net-delay statements (in the INTERCONNECT section of SDF) against the netlist details for the arcs they describe?
(e.g. including the need to traverse through the netlist hierarchy as needed, but this looks like a flat netlist)
This would confirm that indeed there is a consistency mismatch between the two files.
Are you certain that you are not e.g. using an SDF or netlist that may have been for a different step in Encounter?
(e.g. before a final DRC-fixing optimization step, which could insert new cells)
BTW, it is impossible to know what that syntax-error is about without seeing the line it refers to.
(Also, you appear to be seeing some odd bash shell-script errors, but maybe is due to improper coding of your run-script.)
Finally, note that typical modern approach is to not read SDF into PT, but instead to read SPEF or DSPF, since that will annotate caps and resistances using RC-network details so that an "official" static-timing-signoff tool like PT can be the one who performs all of the e.g. Arnoldi delay calculations and so on.
(plus reading in the caps also lets PT check max-tran/max-cap DRC violations without having to read-in a second script of set_load/set_res commands)
But it is also possible that your cell-library/silicon vendor might *require* you to use an SDF using their prescribed tool/flow.