dbms_comparison was introduced in 11g as a powerful tool to compare and synchronize table data in different tables.
Even though it’s quite flexible it has some limitations:
1. The tables are supposed to be in different databases. Therefore if you want to compare tables in the same database you will need to create fake db link.
2. The tables must have unique indexes.

Another consequence from the flexibility is that synchronizing is being applied in two steps (merge & delete) even though it could have been done using single merge statement.
Moreover those statements are based not only on tables that are synchronized but also on auxiliary data populated after comparison.

For the below two tables

create table t1(id, name, constraint pk_t1 primary key (id)) as
  select rownum + 1, 'name src' from dual connect by rownum <= 3;
create table t2(id, name, constraint pk_t2 primary key (id)) as
  select rownum + 2, 'name dst' from dual connect by rownum <= 3;

dbms_comparison will generate two statements for synchronization:

merge into "TEST"."T2"@self t
using (select /*+ USE_NL(d s) */
        s."ID" "ID", s."NAME" "NAME", d.rmt_rowid d_tar_rowid
         from "TEST"."T1" s, "_USER_COMPARISON_ROW_DIF" d
        where d.comparison_id = :cmp_id
          and d.scan_id = :scan_id
          and d.status = 2
          and d.loc_rowid = s.rowid
          and d.idx_val = s."ID") ss
on (t.rowid = ss.d_tar_rowid and t."ID" = ss."ID")
when matched then
  update set t."NAME" = ss."NAME"
when not matched then
  insert ("ID", "NAME") values (ss."ID", ss."NAME")

and

delete "TEST"."T2"@self
 where (rowid, 'X' || "ID") in
       (select d.rmt_rowid, 'X' || d.idx_val
          from "_USER_COMPARISON_ROW_DIF" d
         where d.comparison_id = :cmp_id
           and d.scan_id = :scan_id
           and d.status = 2
           and d.loc_rowid is null)

However it’s possible to merge them like following:

merge into t2
using (select nvl(t2.id, t1.id) id, t1.name, nvl2(t1.id, 1, null) sign
         from t2
         full join t1
           on t1.id = t2.id) t
on (t2.id = t.id)
when matched then
  update
     set t2.name = t.name
   where nvl(t2.name, chr(1)) <> nvl(t.name, chr(1))
  delete
   where t.sign is null
when not matched then
  insert
  values
    (t.id, t.name)

Last statement could be generated dynamically based on tables metadata so such approach remains more preferable for me because allows full control of performance.