diff --git a/bwamem.c b/bwamem.c
index a221d28..6fc2e5a 100644
--- a/bwamem.c
+++ b/bwamem.c
@@ -568,7 +568,7 @@ void mem_chain2aln(const mem_opt_t *opt, int64_t l_pac, const uint8_t *pac, int
 			for (i = 0; i < MAX_BAND_TRY; ++i) {
 				int prev = a->score;
 				aw[0] = opt->w << i;
-				a->score = ksw_extend(s->qbeg, qs, tmp, rs, 5, opt->mat, opt->q, opt->r, aw[0], opt->zdrop, s->len * opt->a, &qle, &tle, &gtle, &gscore, &max_off[0]);
+				a->score = ksw_extend(s->qbeg, qs, tmp, rs, 5, opt->mat, opt->q, opt->r, aw[0], opt->pen_clip, opt->zdrop, s->len * opt->a, &qle, &tle, &gtle, &gscore, &max_off[0]);
 				if (bwa_verbose >= 4) { printf("L\t%d < %d; w=%d; max_off=%d\n", prev, a->score, aw[0], max_off[0]); fflush(stdout); }
 				if (a->score == prev || max_off[0] < (aw[0]>>1) + (aw[0]>>2)) break;
 			}
@@ -591,7 +591,7 @@ void mem_chain2aln(const mem_opt_t *opt, int64_t l_pac, const uint8_t *pac, int
 			for (i = 0; i < MAX_BAND_TRY; ++i) {
 				int prev = a->score;
 				aw[1] = opt->w << i;
-				a->score = ksw_extend(l_query - qe, query + qe, rmax[1] - rmax[0] - re, rseq + re, 5, opt->mat, opt->q, opt->r, aw[1], opt->zdrop, sc0, &qle, &tle, &gtle, &gscore, &max_off[1]);
+				a->score = ksw_extend(l_query - qe, query + qe, rmax[1] - rmax[0] - re, rseq + re, 5, opt->mat, opt->q, opt->r, aw[1], opt->pen_clip, opt->zdrop, sc0, &qle, &tle, &gtle, &gscore, &max_off[1]);
 				if (bwa_verbose >= 4) { printf("R\t%d < %d; w=%d; max_off=%d\n", prev, a->score, aw[1], max_off[1]); fflush(stdout); }
 				if (a->score == prev || max_off[1] < (aw[1]>>1) + (aw[1]>>2)) break;
 			}
diff --git a/ksw.c b/ksw.c
index e331390..2daf809 100644
--- a/ksw.c
+++ b/ksw.c
@@ -359,7 +359,7 @@ typedef struct {
 	int32_t h, e;
 } eh_t;
 
-int ksw_extend(int qlen, const uint8_t *query, int tlen, const uint8_t *target, int m, const int8_t *mat, int gapo, int gape, int w, int zdrop, int h0, int *_qle, int *_tle, int *_gtle, int *_gscore, int *_max_off)
+int ksw_extend(int qlen, const uint8_t *query, int tlen, const uint8_t *target, int m, const int8_t *mat, int gapo, int gape, int w, int end_bonus, int zdrop, int h0, int *_qle, int *_tle, int *_gtle, int *_gscore, int *_max_off)
 {
 	eh_t *eh; // score array
 	int8_t *qp; // query profile
@@ -381,7 +381,7 @@ int ksw_extend(int qlen, const uint8_t *query, int tlen, const uint8_t *target,
 	k = m * m;
 	for (i = 0, max = 0; i < k; ++i) // get the max score
 		max = max > mat[i]? max : mat[i];
-	max_gap = (int)((double)(qlen * max - gapo) / gape + 1.);
+	max_gap = (int)((double)(qlen * max + end_bonus - gapo) / gape + 1.);
 	max_gap = max_gap > 1? max_gap : 1;
 	w = w < max_gap? w : max_gap;
 	// DP loop
diff --git a/ksw.h b/ksw.h
index 2dd6499..97559fd 100644
--- a/ksw.h
+++ b/ksw.h
@@ -102,7 +102,7 @@ extern "C" {
 	 *
 	 * @return        best semi-local alignment score
 	 */
-	int ksw_extend(int qlen, const uint8_t *query, int tlen, const uint8_t *target, int m, const int8_t *mat, int gapo, int gape, int w, int zdrop, int h0, int *qle, int *tle, int *gtle, int *gscore, int *max_off);
+	int ksw_extend(int qlen, const uint8_t *query, int tlen, const uint8_t *target, int m, const int8_t *mat, int gapo, int gape, int w, int end_bonus, int zdrop, int h0, int *qle, int *tle, int *gtle, int *gscore, int *max_off);
 
 #ifdef __cplusplus
 }
diff --git a/main.c b/main.c
index c9acc72..b47c897 100644
--- a/main.c
+++ b/main.c
@@ -3,7 +3,7 @@
 #include "utils.h"
 
 #ifndef PACKAGE_VERSION
-#define PACKAGE_VERSION "0.7.3-r373-beta"
+#define PACKAGE_VERSION "0.7.3-r374-beta"
 #endif
 
 int bwa_fa2pac(int argc, char *argv[]);