diff --git a/apps/readest-app/src/__tests__/utils/txt.test.ts b/apps/readest-app/src/__tests__/utils/txt.test.ts
index 84cca16f..72a4a9d7 100644
--- a/apps/readest-app/src/__tests__/utils/txt.test.ts
+++ b/apps/readest-app/src/__tests__/utils/txt.test.ts
@@ -148,6 +148,41 @@ describe('createChapterRegexps — Chinese (zh) regex matching', () => {
     });
   });
 
+  // Measure-word phrases such as 第一封信 ("the first letter") and 第四本书
+  // ("the fourth book") are running prose, not chapter headings. The volume
+  // measure words 卷/本/册/部/封 only start a heading when followed by a
+  // separator or the line end — not when a noun follows them directly. See
+  // issue #4658.
+  describe('measure-word false positives (issue #4658)', () => {
+    it.each([
+      '第一封信',
+      '第一封信。',
+      '第七封信',
+      '第四本书记载着锤法',
+      '第四本书记载着锤法，来自孙家，可惜对秦铭用处不大了，他已经彻底掌握。',
+      '第三本书',
+      '第十部手机',
+      '第一册书',
+      '第二卷书卷',
+    ])('should NOT match measure-word prose "%s"', (line) => {
+      const regex = getFirstRegex('zh');
+      expect(regex.test(`\n${line}\n`)).toBe(false);
+    });
+
+    it.each([
+      '第一封',
+      '第一封 致读者',
+      '第一本',
+      '第一本 标题',
+      '第二部 中篇',
+      '第一册 上',
+      '第一卷 起始篇',
+    ])('should still match a real heading "%s"', (line) => {
+      const regex = getFirstRegex('zh');
+      expect(regex.test(`\n${line}\n`)).toBe(true);
+    });
+  });
+
   describe('should not match', () => {
     it('should not match chapter heading embedded mid-line', () => {
       const regex = getFirstRegex('zh');
@@ -387,6 +422,35 @@ describe('extractChaptersFromSegment — Chinese (zh)', () => {
     expect(chapters[0]!.content).toContain('前文');
   });
 
+  it('should not extract 第N封信 letter markers as chapters (issue #4658)', () => {
+    const text = [
+      '第一百八十九章 许七安的七封信',
+      '驿卒将信递了过来。',
+      '第一封信',
+      '写这封信的时候我正在云州。',
+      '第二封信',
+      '临安公主亲启。',
+    ].join('\n');
+    const chapters = extractChapters(text, 'zh');
+    const titles = chapters.map((c) => c.title);
+    expect(titles).toEqual(['第一百八十九章 许七安的七封信']);
+    expect(chapters[0]!.content).toContain('第一封信');
+    expect(chapters[0]!.content).toContain('第二封信');
+  });
+
+  it('should not extract 第N本书 + sentence as a chapter (issue #4658)', () => {
+    const text = [
+      '第八十四章 比异人强一点',
+      '他翻阅着那几本秘籍。',
+      '第四本书记载着锤法，来自孙家，可惜对秦铭用处不大了，他已经彻底掌握。',
+      '日子一天天过去。',
+    ].join('\n');
+    const chapters = extractChapters(text, 'zh');
+    const titles = chapters.map((c) => c.title);
+    expect(titles).toEqual(['第八十四章 比异人强一点']);
+    expect(chapters[0]!.content).toContain('第四本书记载着锤法');
+  });
+
   it('should extract 番外 chapters mixed with regular chapters (issue #4016)', () => {
     const text = [
       '第57章 s级冰系掌控',
diff --git a/apps/readest-app/src/utils/txt.ts b/apps/readest-app/src/utils/txt.ts
index a2db46b2..5a543f56 100644
--- a/apps/readest-app/src/utils/txt.ts
+++ b/apps/readest-app/src/utils/txt.ts
@@ -719,15 +719,28 @@ export class TxtToEpubConverter {
     const chapterRegexps: RegExp[] = [];
 
     if (language === 'zh') {
+      // 第N + unit, expressed as two explicit tiers that share the 第N prefix and
+      // the trailing boundary. They stay in ONE alternation (and so one split
+      // pass) on purpose: a segment often mixes chapter and volume headings (a
+      // volume wraps chapters), and the regexps array is a fallback chain — the
+      // first regex that splits "well enough" wins — so separate entries would
+      // recognize one tier and silently drop the other.
+      const cjkNumber = '第[ 　零〇一二三四五六七八九十0-9][ 　零〇一二三四五六七八九十百千万0-9]*';
+      // Tier 1 — chapter units. Real headings; a title may attach directly
+      // (第一章天地初开) or after a separator.
+      const chapterUnit = String.raw`[章节回讲篇话](?:[：:、 　\(\)0-9]*[^\n-]{0,36})`;
+      // Tier 2 — volume/measure-word units. These double as 量词 in prose
+      // (第一封信 "the first letter", 第四本书 "the fourth book"), so a title only
+      // counts when introduced by a separator (：:、, space, parens) or the line
+      // ends — never a bare noun directly after the unit. See issue #4658.
+      const volumeUnit = String.raw`[卷本册部封](?:[：:、 　\(\)][：:、 　\(\)0-9]*[^\n-]{0,36})?`;
+      const numberedHeading = String.raw`${cjkNumber}(?:${chapterUnit}|${volumeUnit})(?!\S)`;
+      const prefaceHeading = String.raw`(?:楔子|前言|简介|引言|序言|序章|总论|概论|后记|番外篇|番外|外传)(?:[：: 　][^\n-]{0,36})?(?!\S)`;
+      const englishHeading = String.raw`chapter[\s.]*[0-9]+(?:[：:. 　]+[^\n-]{0,50})?(?!\S)`;
       chapterRegexps.push(
         new RegExp(
-          String.raw`(?:^|\n)\s*` +
-            '(' +
-            [
-              String.raw`第[ 　零〇一二三四五六七八九十0-9][ 　零〇一二三四五六七八九十百千万0-9]*(?:[章卷节回讲篇封本册部话])(?:[：:、 　\(\)0-9]*[^\n-]{0,36})(?!\S)`,
-              String.raw`(?:楔子|前言|简介|引言|序言|序章|总论|概论|后记|番外篇|番外|外传)(?:[：: 　][^\n-]{0,36})?(?!\S)`,
-              String.raw`chapter[\s.]*[0-9]+(?:[：:. 　]+[^\n-]{0,50})?(?!\S)`,
-            ].join('|') +
+          String.raw`(?:^|\n)\s*(` +
+            [numberedHeading, prefaceHeading, englishHeading].join('|') +
             ')',
           'gui',
         ),