Below is a summary for global (fill-by-fill correlated) uncertainties 
for polarization measurements by CNI group in Run-5,6 and 9 at sqrt(s)=200 GeV.

This is to separate Year-by-Year fully uncorrelated uncertainties and 
correlated ones. The latter one implies to be "non-uncorrelated", 
which could be either fully correlated or partially correlated 
or with unknown correlation. Depending on the particular task 
this "non-uncorrelated" component can be considered as correlated or 
uncorrelated. For example, when averaging asymmetry results from multiple 
years, these should be considered as correlated, 
to follow the coservative or safe approach for uncertainty assignment. 
Uncertainties below are in % (relative):

                 Uncorrel    Correl**  Total
Run5 Blue           5.4       2.4      5.9
Run5 Yell           5.7       2.5      6.2
Run6 Blue           4.0       2.4      4.7
Run6 Yell           4.1       2.5      4.8
Run9 Blue           1.0       3.2      3.3
Run9 Yell           0.8       3.2      3.3

Run5 Blue*Yell      8.1       4.8      9.4
Run6 Blue*Yell      6.8       4.8      8.3
Run9 Blue*Yell      1.3       6.3      6.5

** As mentioned above "Correl" uncertainties may be considered either 
as correlated or uncorrelated depending on particular task.

"Correl" component comes exclusively from H-Jet measurements, used to 
normalize pC measurements. 

Run5 and Run6 values are discribed in detail in: 
http://www4.rcf.bnl.gov/~cnipol/pubdocs/NOTE_2005_2006_Comb_Pol.txt

Run9 values are based on: 
https://wiki.bnl.gov/rhicspin/upload/6/68/Run91112_results.pdf : 

Run9 uncorrelated uncertainties come from pC A_N, 
which was 1.0% for Blue and 0.8 for Yellow. 
Correlated ones include 3% uncertaintiy in jet polarization, 
and 1% uncertainty from background, which are considered as correlated 
between Blue and Yellow beams.

Note, uncertainties break-up in Run9 is quite different from Run5 and Run6, 
due to that different analysis strategies were used in these Years.