Diff in Diff when the treatment time is known but out of the sample range

Hi all,

I am estimating the effects of a shock on industries.
I know the time of the shock, it starts in 2000, but my data on industries cover starts in 2010. If I just estimate a pool dynamic diff-in-diff, everything looks great.
But if I drop observations with a shock before 2010: first, I lose half of my sample. second, my coefficients are not significant anymore. Should I drop these observations? A big part of my sample experienced the shock in 2008...

Thank you
Mahab