Sorry for not working on your pull request. I'm looking for a new maintainer for Python IntelHex project. I hope someone will help.

I understand your intents, but it seems such format is not follow IntelHex-8 standard. Do you have any links or pdf document with hex word format explanations?
Anyway, I don't like the fact you changed the most time-critical function (decode_record). I think it's better to copy decode_record method entirely into new class and make modifications there. Also the IntelHex project is relying on unit tests. We need some unit tests for your class with some simple example of such word file.
Please, rework your class, add unit tests and write some text for manual.

Sorry for not reviewing your patch a year ago. If you don't need it anymore - please say so.

word address mode is described in the last note

I will try to work on fixes.