Probably your definition of baselining needs adjusting. Roddick obviously played well enough from the back to take the advantage over Federer for good chunks of the match. This against a clearly better Federer from the back as well. Roddick's effectiveness off the ground shouldn't be in dispute here, its told in the scoreboard. I don't see how that's any more ludicrous than your position that Federer was only slightly better in 2004...Clarifying Roddick wasn't better per say, just different. Players go for and execute low percentage plays all the time, but its not their usual level hence why its low percentage. Like when Fed served like he was channelling the spirit of Sampras in the 2015 SF, generally Fed ain't a top 10 server but when he's getting that many in and placing it so well he nearly looks like it for a match.
I already said Federer served better in 2015 as well btw. Roddick also executed his gameplan off the ground in excellent fashion for much of the match. The stats are a combined of the two.