Airflow operator's upstream on graph has no upstream list in detail

I have a dependencies list of task ids [['A','B'],['C','D']], which should result in the following stream:

A >> B

C >> D

It generally works except from the case [['A','B']['C','B']], which should results in both operators A and C ending and than executing B. Here is how the dependecies are created:

for dep in dependencies_list:
    for i in range(len(dep)-1):
        task_id_update = "update_bq_ts_{}".format(dep[i])
        task_id_start = "start_df_pipeline_{}".format(dep[i+1])
        last_ops[task_id_update] >> first_ops[task_id_start]

The dictionaries last_ops and first_ops are key:str (task_id) value:Operator.

The B operator (start_df_pipeline_ *) definition looks like this:

def start_pipeline(dag, pipeline_name, runtime_args):
    return CloudDataFusionStartPipelineOperator(
        task_id='start_df_pipeline_{pipeline_name}',
        instance_name=instance_name,
        location=location,
        pipeline_name=pipeline_name,
        runtime_args=runtime_args,
        trigger_rule="all_success"
)

In the graph view it shows the dependecies correctly:

While if I open the Task instance details of strat_df_pipeline_* task the upstream list is empty

And when I run the whole DAG it executes the start_df_pipeline_* not waiting for the two update_bq_* to finish.