autocorrelation

[gobbble]

Hi, I have following regression

Y c b1 b2 d

where
Y is the log return on a price index
b1 is the change in exchange rate
b2 is GDP growth and
d is a dummy variable

the variable b1 shows serial autocorrelation when checking the correlogram.

However, when running the regression the Durbin Watson statistic indicates no autocorrelation which is also confirmed by the Breusch-Godfrey test.

is Autcorrelation a matter or not?

If yes, is the inclusion of ar(1) in the regression the right approach?

Thanks!