Nope.
Their efficiency is almost EXACTLY what the old S7 managed - though the S7 WAS a major improvement over the S5 or even the Spondoolies SP20e running at "efficient" chip voltage.
Bitmain went DIRECTLY to "make it efficient" on the 14/16nm node - they had the Innosilicon A3 (that never ended up shipping apparently because it WASN'T much if any more efficient than the S7) and the first-generation BW.COM SHA256 miners to compete with that were announced products by the time the S9 first shipped, and BitFury had already demoed their 14/16nm chip with BETTER efficiency (at it's most efficient but low hashrate operating point) than the S9 ever managed.
Even with the improvements GF TSMC and Samsung have made to their processes since then, I doubt there is a lot of improvement Bitmain could make to the S9 on efficiency without moving to a new node.
10%, likely.
20%, iffy.
30% - no bloody way at this point.
And given the COST to design chips at 14/16nm, 10% isn't worth their time, 20% probably isn't (though I suspect they've tried some DESIGN work to see if they could get there), and most folks aren't going to consider 30% worth investment into at this late date even if they COULD manage it (or you wouldn't see folks buying Avalon 7xx series and EBang units that are ballpark 30% WORSE than the S9).