The font you are using (SutonnyMJ) is not a valid Unicode font. The Bengali alphabet (abugida) is encoded in the range U+0980-09FF. That particular font has the Bengali characters encoded in the Latin (U+0041-007A) range. Given that the Avro Phonetic keyboard sells itself as being “Full Unicode compliant” (source) I suggest downloading one of the many fonts they provide and see if that still causes problems.
EDIT: Thanks for your feedback. Do you encounter the same problem using iBus>Bengali-Avro Phonetic with other word processing software or only LO? That may help narrow down if it is an LO problem or something related to the keyboard facility being used. If the Unicode fonts they provide do not work, that would tend to indicate that it is an LO problem (IMO). The BiDi characters appear to be getting mis-interpreted.
Another issue might be the DOC format. It shouldn’t be, but it may be worthwhile using an ODT as a test to see if the problem persists. If I save the DOC as and ODT here and then look at the XML, the characters appear fine.
Here is a text file (named with ODT extension) listing the output from the
unicode linux command. It seems like a valid stream from my limited understanding. I also cannot see any byte order mark (BOM) in the octal output from the
od -c content.xml linux command. Even though the Bangla contains a lot of modifying vowel signs, it should display OK.