There isn't much information available for that part of the world. EPSG (http://www.epsg.org - I'm on the subcommittee that maintains it) thinks that the transformation, Arc 1950 to WGS 1984 (3) is incorrect. That Burundi data was on Arc 1960 instead. There's another transformation, Arc 1960 to WGS 1984 (4). You might try defining your data as Arc 1960 with the projection information you have and use that transformation.
None of the transformations that I know about in that area are very good.
Cliff Mugnier in his Grids & Datums column on Burundi lists a GK zone for Burundi that has a scale factor of 1.0, not 0.9996, so that's something else to try.