The reported suggested exchange was much higher than 1:3.5 it was around 1 J-20 defeating many dozens of 4th generation fighters, or in Chinese nomenclature, 3rd generation fighters. I recall the reason of limit given as eventually being overwhelmed but those exchange ratios are allowing the 5th gen to not become saturated and makes lots of assumptions such as returning for refuel and reloading which in war is not always the case. However in perfect world, 5th gen clearly dominated.
This result was the initial tests of J-20 against 4th generations around 2018, reported later around 2020. Then the 4th gen teams developed counters. Basically the whole timeline was like first it was doubted how much more effective 5th gen was by PLAAF and then when they had J-20 in service and ran exercises, they realized the Americans were accurate about how much better 5th gen is. The reports and leaks made it sound like it troubled PLAAF a little bit because before that they still believed air combat is okay to just use 4th gen even against 5th gen. It just required better detection and targeting methods which PLA and PLAAF have invested heavily in ever since F117 was revealed.
I remember one leak then said continuous training against 5th gen made the 4th gen teams develop some better tactics against 5th gen even though the exchange was still heavily in favor of 5th gen. Part of the point of J-20 is also to improve the 4th generation fighters and how they operate against 5th gen fighters. They got the ratio down from 1 to several dozens to something a little bit more favorable but I can't remember if they ever gave a range.
As for J-10C defeating J-16 in last several Golden Helmet exercises, this is suggested as because J-16 RCS is quite large even with some effort to reduce it the engine is exposed. J-10C often detected J-16 well before J-16 found J-10C - said to be over 200km J-10C detected J-16 and took positions to attack them and make better use of terrain and J-16 vectors.
J-10C has quite low RCS actually a lower than Super Hornet and roughly Rafale's. Basically they are small RCS if clean and otherwise when armed it is just the missiles and pylons on its wings that contribute mostly to RCS. J-16 could not find J-10C at over 200km range. J-10C makes use of PL-15 and basically the exponential RCS to detection nature of the fight means J-10C exploit positions and available fuel much more efficiently and would fire first rounds at 200km and then reposition for better second and third rounds of missiles at J-16. J-16 can only realize the fight when those J-10C were found and assumed the missiles are in the air before active to avoid and by then J-10C is in better position and second rounds are on the way and third rounds nearly being fired. Overall that is just enough advantage to ensure J-10C victory overall in enough tries.
Both in close range probably depends totally on the pilot but the J-10C is the king of dogfight in PLAAF I guess after Su-35. J-16 is multirole but more attack fighter than pure dogfighter due to less fuel than single seater flanker and also much more weight than single seater flanker. It also carries a lot of heavy electronic equipment.
These reason make it quite unsurprising J-10C takes the crown during 4th gen Golden Helmet exercises since it participated. Before that, J-10A and J-10B did not fare as well as J-11B and J-16 but J-10C has improved a lot over J-10A in BVR.