In aldol reactions, it is apparently a general rule of thumb that (Z)-enolates give higher stereoselectivity (for the syn aldol product) than (E)-enolates do for the anti product. Quoting Carey & Sundberg, Advanced Organic Chemistry: Part B (5th ed.), p 69:
From these and many related examples the following generalizations can be made about kinetic stereoselection in aldol additions of lithium enolates.
(3) Z-Enolates are more stereoselective than E-enolates.
and the example given (taken from a Clayton Heathcock paper, J. Org. Chem. 1980, 45 (6), 1066–1081) is
Evans also mentions this in his paper comparing the Cornforth vs polar Felkin–Anh models (Angew. Chem. Int. Ed. 2003, 42 (15), 1761–1765):
While Z and E lithium enolates are also formed with good configurational purity, the characteristic decline in syn/anti selectivity is observed, particularly with the E enolate.
Why is this the case? Or, why is $\Delta \Delta G^\ddagger$ for the (Z)-enolate larger (corresponding to greater selectivity) than for the (E)-enolate? I don't see anything that stands out when looking at the six-membered Zimmerman–Traxler transition states - within this model, the difference $\Delta \Delta G^\ddagger$ for both enolate geometries is simply the cost of putting a phenyl group axial instead of equatorial.