For some reason, MemoryMapDataSet is not correctly determining that Unix and Mac file formats use a single byte for newlines (CR or LF). It does DOS format fine (CR and LF). A temporary fix is to convert the file is ConTEXT.
I discovered that the MemoryMapDataSet was reading the wrong bytes when detecting the newline character(s). The FileStream.Seek() call has been corrected and DOS, Unix, and Mac ASCII files are all readable.