In calibre under
Preferences > Common options > Structure detection, the Detect chapters XPath is
Code:
//*[((name()='h1' or name()='h2') and re:test(., '\s*((chapter|book|section|part)\s+)|((prolog|prologue|epilogue)(\s+|$))', 'i')) or @class = 'chapter']
When calibre detects what it thinks is a chapter header, it splits the page at that point. So what's happening is the Kindle eBook is using headers tags for each line. Either h1 and/or h2. So calibre is splitting it as it should.
In order to fix that, you'll have to edit either the XPath or the Kindle eBook. YOu could delete the XPath code, sent the eBook to your Kobo and then put it back.