Odd Consistency in Unemployment Claims Data

(Source: Zero Hedge - Same Trick Different Week: "Initial Claims Decline Following Revision"; Deficit Surge Pushes Q1 GDP To 1.5%)

This data is for the weekly initial and continuing unemployment claims provided by the BLS (Bureau of Labor Statistics). If recollection serves me correctly, the revisions data for 2011 would show a similar, nearly 100%, occurrence of upward revisions. Further, the size of revisions has been relatively stable. Anyone care to explain how an initial estimate for a given statistic, in this case unemployment claims, could consistently miss in the same direction?