The method you tried to use works only for coupling coefficient k close to unity.
I think you could try to estimate the coupling coefficient, to have a more o less accurate measurment of primary, and leakage inductance. I'm going to explain my idea (however I've never tried).
We can refer to the following model:
The transformer is represented as mutual coupling M between a primary having an inductance Lp and a leakage inductance LL, while the secondary has a complessive inductance Ls.
The general equations are:
V1=s*(LL+Lp)*I1+s*M*I2
V2=s*M*I1+s*Ls*I2
then the open secondary (I2=0) inductance (i.e. V1/I1), seen by primary will be:
Lop=LL+Lp
Shorting the secondary, the inductance at the primary will be:
Lsh=LL+Lp*M^2/Ls
Now we can connect the windings in antiseries and measure the inductance again:
In this case I1=-I2 and the inductance (Vx/I1) will be given by:
Zas=LL+Lp+Ls-2*M
Connecting now the winding in series
Here we have I1=I2 thus the inductance Vx/I1 will be:
Zser=LL+Lp-Ls
Now from the system of 4 equations, 4 unknown it's possible to estimate LL, Lp, Ls and M:
Lop=LL+Lp
Lsh=LL+Lp*M^2/Ls
Zas=LL+Lp+Ls-2*M
Zser=LL+Lp-Ls
The coupling coefficient is k≈M/sqrt(Ls*Lp).
I hope this method could works. Let me know if you have any result in applying it.