Yes, you probably do want the cables to be the same length
electrically, especially if you don't have an automatic way to apply a
calibration correction, just for convenience if nothing else. 4
degrees at 1300MHz is about 1.5mm, assuming solid polyethylene
dielectric. You may be able to achieve that sort of matching if you
are really careful in your construction. What sort of connectors are
you using? If they are they type where you solder or crimp the pin to
the center conductor and put the line into the body of the connector
until the dielectric hits the dielectric of the connector, you should
be able to come close to that sort of match in physical length by very
accurately matching the length of the dielectric, and cutting it very
cleanly perpendicular to the axis of the coax. If you're using good
coax, the variation in velocity factor from piece to piece should be
small, and so as long as you don't try to make really long cables, you
should do well.

Note that you can calibrate out lead length differences. A difference
leads presumably to a phase error that ramps linearly with frequency.
Does your VNA have a way to remove/add a time or at least a constant
phase to the readout? And you can perhaps arrange to measure the
electrical length differences. If you can do that, then you can make
up, say, four of the 24" (nom) cables, and pick two matching ones to go
to the two receiver inputs. Note that the length from the source to
the coupler (T/R test set) is NOT CRITICAL at all. So make some cables
as close to the same as you can, and pick the best-matched two for the
receiver cables. That should let you make good reflection
measurements. Then you can either adjust the length of the longer
cable, or calibrate out its length error, for transmission
measurements, assuming you care about phase. Often you'd have to
account for the length of connectors and other cables to your DUT
anyway.

I've made some measurements on some rather expensive test cables and
found the electrical length to vary more than I'd have expected, though
they weren't matched sets. I'd hate to have to select a matched set
from such cables, though, as it would take quite a few cables to assure
that I'd get two that matched close enough! For several hundred
dollars per cable, I'd have expected they'd hold them closer to the
nominal length.

Cheers,
Tom

(Note that some extra care is taken in the design and construction of
the coupler to get accurately matched paths to the two receiver
outputs!)