Whether it's 5% or 50% it's still better. Considering the card is the same or better in other respects there is little sense to go with the 4080 over the 5070 Ti, assuming both are same price. Considering the 4070 Ti has a $250 lower RRP, it's pretty obvious.
The only downside is hassle of returning.
If the difference is so little you would hardly notice or tell then I don't think it really matters. Power Efficiency simply doesn't really matter for this comparison in my opinion. Just because the 7900XTX is a smidge bit faster than the 4080S (usually only by 1% to 4%), suddenly didnt mean people were recommending the 7900XTX over the 4080S in raster. I think the same logic applies here.
Not to mention, OP mentioned overclocking a potential 5070Ti to beat a 4080 (which it already matches as is) which opens up another can of worms for power efficiency, and ATP if your gonna overclock power efficiency probably isnt your concern anymore. And even in this regard, an overclocked 5070Ti from what im reading is still well within 5% difference range. Cooler matters too in that regard but it seems pretty consistent across all the charts I saw for different models (avoiding FE for comparison due to no FE card for 5070Ti)
TL;DR, power efficiency, and by extension, other stuff such as raster, really dont matter in this arguement. I dont see a point in bringing them up. Your choice is gonna entirely hinge on the features. (and price, if you can get a good price of course.)
5070Ti if its at a price similar to a 4080S (which the cheapest ive seen are), if you already own a 4080S, you should pass
unless you value the features exclusive to the 5070Ti and can actually get it. If you don't, and were considering getting a 4080S, then, in that specific instance,
go ahead. Even more so if you can buy a 5070Ti at its
actual MSRP. If your not an american thats a factor too, due to stuff like VAT, but thats getting into a whole other can of worms.
