diff --git a/align.c b/align.c
index 277377c..d582c6e 100644
--- a/align.c
+++ b/align.c
@@ -200,7 +200,7 @@ static void mm_align_pair(void *km, const mm_mapopt_t *opt, int qlen, const uint
 	if (opt->flag & MM_F_SPLICE)
 		ksw_exts2_sse(km, qlen, qseq, tlen, tseq, 5, mat, opt->q, opt->e, opt->q2, opt->noncan, zdrop, flag, ez);
 	else if (opt->q == opt->q2 && opt->e == opt->e2)
-		ksw_extz2_sse(km, qlen, qseq, tlen, tseq, 5, mat, opt->q, opt->e, w, zdrop, flag, ez);
+		ksw_extz2_sse(km, qlen, qseq, tlen, tseq, 5, mat, opt->q, opt->e, w, zdrop, end_bonus, flag, ez);
 	else
 		ksw_extd2_sse(km, qlen, qseq, tlen, tseq, 5, mat, opt->q, opt->e, opt->q2, opt->e2, w, zdrop, end_bonus, flag, ez);
 }
diff --git a/chain.c b/chain.c
index 4919b51..77c65b2 100644
--- a/chain.c
+++ b/chain.c
@@ -21,8 +21,8 @@ static inline int ilog2_32(uint32_t v)
 
 mm128_t *mm_chain_dp(int max_dist_x, int max_dist_y, int bw, int max_skip, int min_cnt, int min_sc, int is_cdna, int n_segs, int64_t n, mm128_t *a, int *n_u_, uint64_t **_u, void *km)
 { // TODO: make sure this works when n has more than 32 bits
-	int32_t st = 0, k, *f, *p, *t, *v, n_u, n_v;
-	int64_t i, j;
+	int32_t k, *f, *p, *t, *v, n_u, n_v;
+	int64_t i, j, st = 0;
 	uint64_t *u, *u2, sum_qspan = 0;
 	float avg_qspan;
 	mm128_t *b, *w;
diff --git a/ksw2.h b/ksw2.h
index 43047c2..e82dd82 100644
--- a/ksw2.h
+++ b/ksw2.h
@@ -51,7 +51,7 @@ void ksw_extz(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t
 			  int8_t q, int8_t e, int w, int zdrop, int flag, ksw_extz_t *ez);
 
 void ksw_extz2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat,
-				   int8_t q, int8_t e, int w, int zdrop, int flag, ksw_extz_t *ez);
+				   int8_t q, int8_t e, int w, int zdrop, int end_bonus, int flag, ksw_extz_t *ez);
 
 void ksw_extd(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat,
 			  int8_t gapo, int8_t gape, int8_t gapo2, int8_t gape2, int w, int zdrop, int flag, ksw_extz_t *ez);
diff --git a/ksw2_dispatch.c b/ksw2_dispatch.c
index 56fbd63..c4e3e07 100644
--- a/ksw2_dispatch.c
+++ b/ksw2_dispatch.c
@@ -50,16 +50,16 @@ int x86_simd(void)
 	return flag;
 }
 
-void ksw_extz2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat, int8_t q, int8_t e, int w, int zdrop, int flag, ksw_extz_t *ez)
+void ksw_extz2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat, int8_t q, int8_t e, int w, int zdrop, int end_bonus, int flag, ksw_extz_t *ez)
 {
-	extern void ksw_extz2_sse2(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat, int8_t q, int8_t e, int w, int zdrop, int flag, ksw_extz_t *ez);
-	extern void ksw_extz2_sse41(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat, int8_t q, int8_t e, int w, int zdrop, int flag, ksw_extz_t *ez);
+	extern void ksw_extz2_sse2(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat, int8_t q, int8_t e, int w, int zdrop, int end_bonus, int flag, ksw_extz_t *ez);
+	extern void ksw_extz2_sse41(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat, int8_t q, int8_t e, int w, int zdrop, int end_bonus, int flag, ksw_extz_t *ez);
 	unsigned simd;
 	simd = x86_simd();
 	if (simd & SIMD_SSE4_1)
-		ksw_extz2_sse41(km, qlen, query, tlen, target, m, mat, q, e, w, zdrop, flag, ez);
+		ksw_extz2_sse41(km, qlen, query, tlen, target, m, mat, q, e, w, zdrop, end_bonus, flag, ez);
 	else if (simd & SIMD_SSE2)
-		ksw_extz2_sse2(km, qlen, query, tlen, target, m, mat, q, e, w, zdrop, flag, ez);
+		ksw_extz2_sse2(km, qlen, query, tlen, target, m, mat, q, e, w, zdrop, end_bonus, flag, ez);
 	else abort();
 }
 
diff --git a/ksw2_extz2_sse.c b/ksw2_extz2_sse.c
index e8f71a4..c0671a1 100644
--- a/ksw2_extz2_sse.c
+++ b/ksw2_extz2_sse.c
@@ -15,12 +15,12 @@
 
 #ifdef KSW_CPU_DISPATCH
 #ifdef __SSE4_1__
-void ksw_extz2_sse41(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat, int8_t q, int8_t e, int w, int zdrop, int flag, ksw_extz_t *ez)
+void ksw_extz2_sse41(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat, int8_t q, int8_t e, int w, int zdrop, int end_bonus, int flag, ksw_extz_t *ez)
 #else
-void ksw_extz2_sse2(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat, int8_t q, int8_t e, int w, int zdrop, int flag, ksw_extz_t *ez)
+void ksw_extz2_sse2(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat, int8_t q, int8_t e, int w, int zdrop, int end_bonus, int flag, ksw_extz_t *ez)
 #endif
 #else
-void ksw_extz2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat, int8_t q, int8_t e, int w, int zdrop, int flag, ksw_extz_t *ez)
+void ksw_extz2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat, int8_t q, int8_t e, int w, int zdrop, int end_bonus, int flag, ksw_extz_t *ez)
 #endif // ~KSW_CPU_DISPATCH
 {
 #define __dp_code_block1 \
@@ -291,10 +291,14 @@ void ksw_extz2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 	if (!approx_max) kfree(km, H);
 	if (with_cigar) { // backtrack
 		int rev_cigar = !!(flag & KSW_EZ_REV_CIGAR);
-		if (!ez->zdropped && !(flag&KSW_EZ_EXTZ_ONLY))
+		if (!ez->zdropped && !(flag&KSW_EZ_EXTZ_ONLY)) {
 			ksw_backtrack(km, 1, rev_cigar, 0, (uint8_t*)p, off, off_end, n_col_*16, tlen-1, qlen-1, &ez->m_cigar, &ez->n_cigar, &ez->cigar);
-		else if (ez->max_t >= 0 && ez->max_q >= 0)
+		} else if (!ez->zdropped && (flag&KSW_EZ_EXTZ_ONLY) && ez->mqe + end_bonus > ez->max) {
+			ez->reach_end = 1;
+			ksw_backtrack(km, 1, rev_cigar, 0, (uint8_t*)p, off, off_end, n_col_*16, ez->mqe_t, qlen-1, &ez->m_cigar, &ez->n_cigar, &ez->cigar);
+		} else if (ez->max_t >= 0 && ez->max_q >= 0) {
 			ksw_backtrack(km, 1, rev_cigar, 0, (uint8_t*)p, off, off_end, n_col_*16, ez->max_t, ez->max_q, &ez->m_cigar, &ez->n_cigar, &ez->cigar);
+		}
 		kfree(km, mem2); kfree(km, off);
 	}
 }
diff --git a/main.c b/main.c
index 554ac48..b2349f0 100644
--- a/main.c
+++ b/main.c
@@ -6,7 +6,7 @@
 #include "mmpriv.h"
 #include "getopt.h"
 
-#define MM_VERSION "2.2-r500-dirty"
+#define MM_VERSION "2.2-r501-dirty"
 
 #ifdef __linux__
 #include <sys/resource.h>