Of course Mathematica is more succinct - it is a domain specific language, not general purpose like most of the others compared. The only meaningful comparisons to Mathematic would be R and MATLAB (other DSLs).
That I think is owed to Mathematica being an heir of LISP that implemented m-expressions so cleanly that M doesn't even feel like a homoiconic language. Most other LISP-derivates (like Clojure) stuck to s-expressions, which are historical baggage more than anything else. Some day I'm sure we'll get an m-expression version of Clojure.
Of course in practice one defines new DSLs quite often in Mathematica. And it comes with a variety of DSLs "built-in", mainly for user interface construction, vector graphics, mathematics, and statistical computations.
Disclosure: I too work for WRI.
See this[0] for more info
[0] http://reference.wolfram.com/legacy/v5/Tour/MathematicaAsAPr...
There is no reason to think this looking at the website.
Since he uses median ratios, then a single spectacular ratio will not skew the results.
The part of the test set that people have actually provided Mathematica solutions for. There is every reason to believe that people are more likely to provide Mathematica solutions for problems that Mathematica is likely to be good at.
Most of these apps come from people in the target market, but it's similar to the number of Excel spreadsheets that end up running surprisingly large companies. A "non-programmer" develops a bunch of macros and VBA code to make their job easier, and those "DSL" type capabilities become critical core aspects of the system.