Welcome, @claudiaq - would be helpful if you could post the
inp file as a dropbox link or equivalent. That way we could run the system and discuss the particulars. But in general, water age can take quite some time to evolve into a stable range at the extremities of the system - depends on tank retention time, dead-end demands, etc.
The initial "burn-in" can sometimes be reduced by setting a global water age as an initial condition, but some allowance should be made for that initial period of evolution regardless. Ideally should account for the total system retention including effects of tank mixing. In practice, I've seen some networks that need several weeks of simulation backfill to reach "stability" in water age predictions.