While one of your two "most studies" says TV debates influence the final decisions of "only" 2-3% of voters, the other one says: 3% of voters surveyed said that what they hear in debates will very likely impact their eventual choice and an additional 10% said it will somewhat likely impact their final choice.

Considering that most US presidential elections are decided on very small margins, e.g. from 2000 till 2020, the winning percentage in votes have been between 0.5% and 4.5% with the exception of Obama in 2008 which was about 7% (still less than 10%), I think debates still have a significant impact on election results.

Scrutinising your main source (the research paper by Le Pennec & Pons) further, one could even argue that their findings have little or no bearing on the current situation in the US. All the debates they researched happened less than 45 days to their respective elections, with the average being 24 days. Also, the incumbent this time didn't just have a "bad day", the whole world saw what it was (of course the partisans will always be in denial of this fact).

Cheers!