Corruption in the BEA's initial GDP release?

I converted each GDP data componenet for 2006 Q3 into a standard deviation (compared with the data series volatility over the last 3 years). This is useful for finding components which had large changes, but might be getting hidden because they make up a small part of the overall GDP number. I found something surprising.

First the results:

Top 5 increases:
Nondurable goods - Energy: 1.7 std deviations
Agenda - Motor vehicle output: 1.7 std deviations
State and local consumption expenditures: 1.6 std deviations
Nonredidential fixed structures: 1.5 std deviations
Durable goods - Motor vehicles and parts: 0.7 std deviations

Top 5 decreases:
GPDI Residential: -3.0 std deviations
Agenda - Structures: -2.2 std deviations
Nondurable goods - Food: -2.0 std deviations
Gross private domestic investment: -1.2 std deviations
Nondurable goods - Other: -1.0 std deviations

Graphically:



I checked in 3q 2002, 3q 1998, 3q 1992, 3q 1988, and 3q 1984 and all EXCEPT 3q 1988 had top 5 increase appearances by both Agenda - motor vehicle output AND state and local consumption expenditures. These data series were then revised down by at least 1/2 of a standard deviation in 8 out of the 9 cases (2002 did not have a downward revision in state and local consumption expenditures).

Sorry, BEA, your data has been corrupted by members of congress.