From the research paper linked from the article, here are how the kids scored in different contexts:

  calculating prices in the street      98%
  solving word problems on an exam      74%
  solving abstract problems on an exam  34%

This would suggest that applying concepts is more important, physical context less so.

Mind you, this is an extremely difficult experiment to carry out scientifically and is open to many interpretations. I would at least want to see it reproduced before giving it much consideration.