1. TO_DATE 함수 사용 시 포멧 지정
 - 포멧 지정 안하면 오류 가능성 높습니다.
 - 포멧을 꼭 지정해 주세요.
2. 일수 계산은
 - e_day - s_day + 1
 

WITH tech AS
(
SELECT 'AAA' cd, '002' itsqf, '20200101' s_day, '20201231' e_day FROM dual --  1
UNION ALL SELECT 'AAA', '005', '20200201', '20201231' FROM dual --  0
UNION ALL SELECT 'BBB', '001', '20060101', '20201231' FROM dual -- 15
UNION ALL SELECT 'BBB', '002', '20200201', '20201231' FROM dual --  0
UNION ALL SELECT 'BBB', '003', '20200201', '20201231' FROM dual --  0
UNION ALL SELECT 'BBB', '004', '19980101', '20201231' FROM dual -- 23
UNION ALL SELECT 'CCC', '001', '20200201', '20201231' FROM dual --  0
UNION ALL SELECT 'CCC', '003', '20200201', '20201231' FROM dual --  0
UNION ALL SELECT 'CCC', '004', '20200201', '20201231' FROM dual --  0
UNION ALL SELECT 'DDD', '002', '20050101', '20201231' FROM dual -- 16
UNION ALL SELECT 'DDD', '005', '20200201', '20201231' FROM dual --  0
UNION ALL SELECT 'EEE', '003', '20180101', '20201231' FROM dual --  3
UNION ALL SELECT 'FFF', '001', '20200101', '20201231' FROM dual --  1
UNION ALL SELECT 'FFF', '004', '20200201', '20201231' FROM dual --  0
UNION ALL SELECT 'GGG', '001', '20180101', '20201231' FROM dual --  3
)
SELECT NVL(a.y_nm, '합계') y_nm
     , COUNT(DECODE(b.itsqf, '001', 1)) "001"
     , COUNT(DECODE(b.itsqf, '002', 1)) "002"
     , COUNT(DECODE(b.itsqf, '003', 1)) "003"
     , COUNT(DECODE(b.itsqf, '004', 1)) "004"
     , COUNT(DECODE(b.itsqf, '005', 1)) "005"
  FROM (SELECT LEVEL - 1 y
             , CASE WHEN LEVEL = 1 THEN '1년 미만'
                    WHEN LEVEL = 6 THEN '5년 이상'
                    ELSE (LEVEL-1) || '년' END y_nm
          FROM dual
         CONNECT BY LEVEL <= 6
        ) a
     , (SELECT cd
             , itsqf
             , LEAST(5, FLOOR((TO_DATE(e_day, 'yyyymmdd') - TO_DATE(s_day, 'yyyymmdd') + 1) / 365)) y
          FROM tech
        ) b
 WHERE a.y = b.y(+)
 GROUP BY ROLLUP((a.y, a.y_nm))
 ORDER BY a.y
;

왜 그래야 하는지? 설명이 없네요?
한참 고민했습니다. itsqf 별로 1차 집계가 필요해 보입니다.
출력 기준을 1년 줄였네요? 그러면 LEAST 함수의 인수도 수정이 필요합니다.

WITH tech AS
(
SELECT 'AAA' cd, '001' itsqf, '20200131' s_day, '20210103' e_day FROM dual --  0
UNION ALL SELECT 'AAA', '001', '20190101', '20210103' FROM dual --  2
UNION ALL SELECT 'BBB', '001', '20200105', '20210120' FROM dual --  1
)
SELECT NVL(a.y_nm, '합계') y_nm
     , COUNT(DECODE(b.itsqf, '001', 1)) "001"
  FROM (SELECT LEVEL - 1 y
             , CASE WHEN LEVEL = 1 THEN '1년 미만'
                    WHEN LEVEL = 5 THEN '4년 이상'
                    ELSE (LEVEL-1) || '년' END y_nm
          FROM dual
         CONNECT BY LEVEL <= 5
        ) a
     , (SELECT cd
             , itsqf
--           , LEAST(5, FLOOR(   (TO_DATE(e_day, 'yyyymmdd') - TO_DATE(s_day, 'yyyymmdd') + 1) / 365)) y
             , LEAST(4, FLOOR(SUM(TO_DATE(e_day, 'yyyymmdd') - TO_DATE(s_day, 'yyyymmdd') + 1) / 365)) y
          FROM tech
         GROUP BY cd, itsqf    -- 그룹바이 및 SUM 추가
        ) b
 WHERE a.y = b.y(+)
 GROUP BY ROLLUP((a.y, a.y_nm))
 ORDER BY a.y
;