Panel Estimation with GLS Cross Section Weights

[abcxyz]

In estimating Panel Regression equation, I found nearly all the variables to be insignificant but when I used GLS CROSS SECTION WEIGHTS in Panel Options, then the story changed completely and most of the variables became significant but still the Durbin Watson statistics has changed from 1.12 to 1.36.

What does this imply ?? Heteroscedaticity or/and Autocorrelation ?? Or Something else.

I would be obliged if anybody faced with the same problem let me know what is the situation like.

Regards.