density loss in 1d-advection.Why? coarse mesh? velocity field?

Dear all,
as suggested in this forum many times, I used CLAWPACK for solving a pde, using the clawpack "APPLICATION" directory, where many common pde are solved.
My question is: is the solution strictly conservative? I am measuring loss in the integral of the pulse that I am sending (a loss of the 19.9%!!) even with Courant numbers of .99.
The velocity is spatially varying, but I think very slowly: in a single time step the max coeff of variation is 0.198 and max range of 0.0064 and in this test velocity is changing very smoothly.
There should not be any particular code bug and with a simple 1st ord upwind I have similar losses (20.1%).
It might be a problem of grid? Now I am using a mesh of 20m for a linear lenght of 480m (N=24 cells): I am asking this in advance, because to test a finer grid is absolutely not an immediate task in my code...
Thanks for the help!

A
.