The K13 coordinates plot OK on the PCA, but in the admixture runs they are not directly compatible with regular G25, because they rather pick up other K13 converts than G25 regulars and the same is true vice versa.
However, among themselves, these K13 coordinates seem to be comparable and they plot reasonably well (not perfect) on the PCA.
I therefore made first a custom PCA for the Romanian samples:
One can see the dense cluster for the Romanian core and those which deviate in a Southern (Moldova Gagauzia and Tighina), Slavic (Moldova Balti), Hungarian (Maramures) and German (Satu Mare) direction.
In the centre of the Romanian core is clearly Oltenia in my opinion.
The next thing I did was using Oltenia as the core group and all the main minorities and neighbours as additional references for an admixture run - of course they are not all ideal, but like I wrote before, regular G25 don't get picked up as much. With regular G25 one could use German Erlangen, Hungarian, Russian Kursk and Greek Laconia as proxies, but because of the different original formats, that doesn't work as well.
This was the result of the admixture run sorted by Oltenia-like percentages:
For some models it causes troubles to differentiate between Slavic, Hungarian and German, because Hungarian is to large degree a mix of German and Slavic, and even the Southern admixture is similar to Romanians.
Anyway, I think its a nice approach and the Oltenia sample is really a good representative for core Romanian, with many other regions showing varying degrees of admixture from the neighbours and minorities.