In reply to cgales:
Yes, you have identified the issue, without cloning, the items get emptied right after the write() call. In fact the reason why I have commented out the line where the items get cloned, is because I have seen that the cloned items are always empty, while the items themselves are not (after the cloning), so I have decided to use the items themselves. But after debugging I found out why cloning does not work, it is because I have the do_copy() function commented out, I thought that obj.clone() does not use this function, but it does. Now my items are observed in the scoreboard and it has nothing to do with the uvm_tlm_analysis_fifo TLM connection (which I was suspecting in the beginning).
Thanks