To me, the short answer would be SRS, but everything depends on the specific numbers.
Let's say Team A has the best record in the league and Team B has the best SRS, in four different scenarios.
In scenario 3, the disparity in records is pretty huge, but the SRS is close to identical. The opposite is true in scenario 4. So in either case, "which is the better metric?" kind of becomes an irrelevant question because the only data there that shows any significant difference is one or the other. So even though I think SRS is in fact better, in scenario 3 I think it would be totally reasonable to predict that Team A would have greater playoff success (despite slightly trailing Team B in SRS) based on the big gap in their records. (This of course assuming both teams were healthy all season, etc.).
Team A 65-18 5
Team B 60-23 8
Team A 63-20 6
Team B 61-22 7
Team A 68-15 6
Team B 58-25 7
Team A 63-20 4
Team B 61-22 9
In scenario 2, the difference between the two teams in both categories is so slight that nothing meaningful could be concluded.
But what we're really talking about is scenario 1, where there's (what could be seen as - I don't know if others here will agree) a comparable difference between record and SRS, big enough to be significant, but the record advantage going to Team A, and SRS going to Team B. And I'm pretty sure the data is there to suggest that SRS will be the more reliable predictor in such cases. (Wasn't that what Chronz's big post was all about?). So in scenario 1, Team B would probably have the edge based on a stronger SRS, despite the worse record.
By "Exp W-L Ratio" do you mean Pythagorean wins/losses? That's the only thing along those lines that I've seen people use in the way you're talking about.