I see, the schema's are a bit small, but I think I can make it out. You are saying the first image is the one to implement, and then some power supply filtering...?
The original instructions for the original board can be found at
http://115.28.16.44:81/file/3156.doc (also
https://kechuang.org/t/62913 )
Youtube videos on direct sampling (all of them)
https://www.youtube.com/results?search_query=direct+samplingThe Ba5SBA 'mother board" uses a single filter designed to cut off at 28 MHz, but really, two separate filters - 0-14 MHz and 14-28 Mhz would likely work much better.
Also, the input impedance of the RTL chip input is likely much higher than the several hundred ohms suggested by the use of a 4:1 impedance transformation. from 50 .. So a higher ratio of turns in the transformer might be helpful.. if the antenna was 50 ohms..
See discussion at
http://radiochief.ru/kv-ukv/sobiraem-shirokopolosnyj-sdr-priemnik-na-baze-rtl2832u/Maybe an antenna matching circuit would be good there?
It would likely be worth it to see if that 1:2 ratio was wrong.. at least for your antenna.
Also, I would try other transformer materials than 43 - Others would be better for much of HF..
I would also try a binocular core or perhaps multiple balun cores..
This is an investigation which other more knowledgeable people here could add a lot more to it than I can.