The Alb_mdv samples are straight out of the papers dataset as Lazaridis labeled them, it includes all of them.
Test them out on G25 outside of some outliers they indeed show low Slavic component vis a vis modern Albanians. I used Alb_PostMdv because the only modern (south)Albanian reference in the dataset looks like an outlier. But once Reich lab release the updated HO geno file even this wont be an issue as it has more references.
No, fstats does not work like G25 where you can get good distances with even nonsensical, ahistorical components. It is very hard to make viable models, and if I understand it correctly the models have a 95% confidence interval.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8049561/
In G25 you might be able to do Albanian = Mycenean + loads of Slavic as you say, but not here:
These stuff just gets rejected.
Edit:
^Neither do the models make sense, neither are they viable
Now
^This one is close but still doesn't work, cause as I said the main reference in the dataset for modern Albanians acts weird. Lets try mdvAlb maybe it will work
Went through the edit trouble, now if you have any integrity admit that you are wrong.
PS: By Thracians I assume you mean BGR ancient samples, well you should know that they are a lot closer autosomaly to Myceneans than the Macedonian samples. If anything Macedonian samples cluster as Illyrians and Cetina. Hence even if the very last model worked it would not have worked the way you thought.
Naming and visualizing the clusters: