diff --git a/hit.c b/hit.c
index 25cf538..b6585d1 100644
--- a/hit.c
+++ b/hit.c
@@ -79,6 +79,35 @@ void mm_set_parent(void *km, float mask_level, int n, mm_reg1_t *r) // and compu
 	kfree(km, w);
 }
 
+void mm_update_parent(void *km, float mask_level, int n, mm_reg1_t *r) // due to changes to r.{qs,qe} after DP extension
+{
+	int i, j, k, *w, n_pri = 0;
+	if (n <= 0) return;
+	for (i = 0; i < n; ++i)
+		if (r[i].id == r[i].parent) ++n_pri;
+	if (n_pri <= 1) return;
+	w = (int*)kmalloc(km, n_pri * sizeof(int));
+	for (i = j = 0; i < n; ++i) // find the first primary
+		if (r[i].id == r[i].parent) break;
+	for (w[0] = i, i = i + 1, k = 1; i < n; ++i) {
+		int si = r[i].qs, ei = r[i].qe;
+		if (r[i].id != r[i].parent) continue; // only check primary
+		for (j = 0; j < k; ++j) {
+			int sj = r[w[j]].qs, ej = r[w[j]].qe;
+			int min = ej - sj < ei - si? ej - sj : ei - si;
+			int ol = si < sj? (ei < sj? 0 : ei < ej? ei - sj : ej - sj) : (ej < si? 0 : ej < ei? ej - si : ei - si);
+			if (ol > mask_level * min) {
+				r[i].parent = r[w[j]].parent;
+				if (r[w[j]].subsc < r[i].score)
+					r[w[j]].subsc = r[i].score;
+				break;
+			}
+		}
+		if (j == k) w[k++] = i;
+	}
+	kfree(km, w);
+}
+
 void mm_sync_regs(void *km, int n_regs, mm_reg1_t *regs) // keep mm_reg1_t::{id,parent} in sync; also reset id
 {
 	int *tmp, i, max_id = -1, n_tmp, n_pri;
@@ -109,7 +138,6 @@ void mm_select_sub(void *km, float mask_level, float pri_ratio, int *n_, mm_reg1
 {
 	if (pri_ratio > 0.0f && *n_ > 0) {
 		int i, k, n = *n_;
-		mm_set_parent(km, mask_level, n, r);
 		for (i = k = 0; i < n; ++i)
 			if (r[i].parent == i || r[i].score >= r[r[i].parent].score * pri_ratio)
 				r[k++] = r[i];
diff --git a/main.c b/main.c
index d24c952..dfe4ddc 100644
--- a/main.c
+++ b/main.c
@@ -10,7 +10,7 @@
 #include "minimap.h"
 #include "mmpriv.h"
 
-#define MM_VERSION "2.0-r121-pre"
+#define MM_VERSION "2.0-r122-pre"
 
 void liftrlimit()
 {
diff --git a/map.c b/map.c
index 1ec5fec..1f4f9c9 100644
--- a/map.c
+++ b/map.c
@@ -238,11 +238,17 @@ mm_reg1_t *mm_map_frag(const mm_mapopt_t *opt, const mm_idx_t *mi, mm_tbuf_t *b,
 	regs = mm_gen_regs(b->km, qlen, n_u, u, a);
 	*n_regs = n_u;
 	if (!(opt->flag & MM_F_AVA)) { // don't choose primary mapping(s) for read overlap
+		mm_set_parent(b->km, opt->mask_level, *n_regs, regs);
 		mm_select_sub(b->km, opt->mask_level, opt->pri_ratio, n_regs, regs);
 		mm_join_long(b->km, opt, qlen, *n_regs, regs, a); // TODO: this can be applied to all-vs-all in principle
 	}
-	if (opt->flag & MM_F_CIGAR)
+	if (opt->flag & MM_F_CIGAR) {
 		regs = mm_align_skeleton(b->km, opt, mi, qlen, seq, n_regs, regs, a);
+		if (!(opt->flag & MM_F_AVA)) {
+			mm_update_parent(b->km, opt->mask_level, *n_regs, regs);
+			mm_select_sub(b->km, opt->mask_level, opt->pri_ratio, n_regs, regs);
+		}
+	}
 	mm_set_mapq(*n_regs, regs);
 
 	// free
diff --git a/mmpriv.h b/mmpriv.h
index 65f82a1..bcbb3c7 100644
--- a/mmpriv.h
+++ b/mmpriv.h
@@ -39,6 +39,8 @@ mm_reg1_t *mm_align_skeleton(void *km, const mm_mapopt_t *opt, const mm_idx_t *m
 mm_reg1_t *mm_gen_regs(void *km, int qlen, int n_u, uint64_t *u, mm128_t *a);
 void mm_split_reg(mm_reg1_t *r, mm_reg1_t *r2, int n, int qlen, mm128_t *a);
 void mm_sync_regs(void *km, int n_regs, mm_reg1_t *regs);
+void mm_set_parent(void *km, float mask_level, int n, mm_reg1_t *r);
+void mm_update_parent(void *km, float mask_level, int n, mm_reg1_t *r);
 void mm_select_sub(void *km, float mask_level, float pri_ratio, int *n_, mm_reg1_t *r);
 void mm_join_long(void *km, const mm_mapopt_t *opt, int qlen, int n_regs, mm_reg1_t *regs, mm128_t *a);
 void mm_set_mapq(int n_regs, mm_reg1_t *regs);