Ahhh, got it. I've no indication of peoples' level of knowledge here other than what's immediately mentioned (or occasionally remembered from prior conversation) so I tend to include such details in case they are relevant, and conversely are superfluous when not, which is fine.
So what's missing? Make a resistance bridge, perhaps?
Like I said, I would test it by wiring up terminations and measuring pairs of ports. Prove that the gain between ports is as it's supposed to be (one port (normal mode) goes into +/-1/2 at CM/DM ports; CM/DM go into +/-1/2 (normal) at the balanced ports), CM is easy enough to measure (set up a 0° splitter into the balanced ports), and, that doesn't completely constrain what DM gain/phase can be but it greatly narrows it down. Just going through the permutations of ports should be illuminating. Preferably this would be done as a vector measurement, but even doing magnitude on the spec or even scope is something.
You could further test the balance by making a 180° splitter; this could be as simple as a 0° splitter into an inverter (1:1 transformer), plus a TL to balance the delay. You could for example take two equal lengths of coax, wire one to the splitter (compensation delay, nothing else), and wire the other to the splitter but cut it in half in the middle to swap +/-, and stack ferrite cores all along it (or wind it around a core). The outputs of which can be tested with a resistor divider from one port to the other, the midpoint of which should read zero (zero CM = zero phase/amplitude error in the balun; there could still be a differential gain error where both vary proportionally but this can be constrained by measuring each port normal-mode).
Or you could construct two DM/CM splitters and verify their cascaded balance. See: measuring one antenna by building two and measuring their combined transfer function.
Tim