that's fascinating, I thought it would be symetric enough to be basically really hard to measure. 5% really surprised me.
What does this, like a non isotropic core and coupling to nearby parts?
Maybe it has to do with the test fixture.. you would imagine if its on a bare PCB away from anything, 5 % seems crazy for machine made or film parts.
I heard about this phenomena before and I thought maybe 0.5% max error