About the second case, it was likely due to time spent on matching unicode digits, as per the reasoning for this stackoverflow post,

http://stackoverflow.com/questions/16621738/d-is-less-effici...