select -- Выбираем состояния на последний день периода:
       dt x_dt
     , max(cnt) keep(dense_rank last order by dto, lv) as x_cnt
     , max(sum_sm) keep(dense_rank last order by dto, lv) as x_sm
  from ( -- По сути, делаем расчет "эрланга":
         select dt, dto, lv
              , sum(cnt) over(order by dto,lv) as cnt
              , sum(sm) over(order by dto,lv) as sum_sm
           from (-- Разворачиваем диапазоны в точки смены состояний:
                  select decode(v.lv,1,dts,dtf) dto
                        , decode(v.lv,1,decode(dts,trunc(dts,'mm'),dts,last_day(trunc(dts,'mm') )+1)
                                       ,decode(dtf,trunc(dtf,'mm'),dtf,last_day(trunc(dtf,'mm') )+1)
                                ) as dt
                        , decode(v.lv,1,sm,-sm) sm
                        , decode(v.lv,1,1,-1) cnt, v.lv
                   from tab 
                   -- "Опорная" табличка на две строки, для разворота TAB
                  cross join (select level as lv from dual connect by level <= 2) v
                )
       )
 group by dt
 order by dt;