Creating a loading job results in hardcoded vertices on the running loading job

I was trying to track down an error where an edge file csv was somehow creating vertices; I found that when I created a loading job with:

CREATE LOADING JOB e_associated_with_v1 FOR GRAPH ray {\nDEFINE FILENAME e_associated_with = “/e_associated_with_v1.csv”;\n LOAD e_associated_with TO EDGE associated_with VALUES($0 $1, $2 $3, $4, $5) USING SEPARATOR=",", HEADER=“true”, EOL="\n", QUOTE=“double”;\n

The $1 was showing up as a hard-coded vertex when I do an ‘ls’ and look at the loading job in gsql.
VALUES($0 ip, $2 $3, $4, $5)

This causes an error. We are on an old version 3.1 (upgrading in progress), but hoping to get past this–why would that happen? Some kind of optimization? Thanks.

The answer to this is that even with undirected edges, the order specified in the schema dictates order. The option VERTEX_MUST_EXIST can be useful for preventing unintended vertices, but doesn’t solve the order issue. Thanks.