Got this from a guy on Facebook epub production page,
"I had similar problems in other programs. Typically it is caused by the wrong encoding. Each character can be expressed by more than one entity. Unfortunately, not all applications can interpret every entity type. Some entities are all numbers, such ӄ while others can be expressed with letters, such as You will find that most applications support the first, but not all support the second. All letters and character also have a hex code. Sometimes the solution is simply using the supported entity, in others you have to declare the encoding at the top of the document. Also, be aware that “plain text” documents may not support non-ASCII character sets. In 1995, most computer apps and file types switched to Unicode which supports all TrueType character sets. But you will have to check to see if the app you’re using supports ASCII or Unicode. I believe ASCII supports only 256 characters. I had many of these issues back when I was working with XML."