First of all, thank you. I value rigorous evaluation of methods and if there is robust counter-evidence, it should not be taken lightly.

I read your post and was unable to find any reproducible code that supports your claims. I would love to see any of that work if possible.