If the coupon is welded propely, there will be some melt-through on the root side. The line will run right down the middle of the root reinforcement. That area is thicker than the adjacent base metal, thus, complete joint penetration is achieved and the is no incomplete fusion. If however, the welder does not preclean the parts completely, i.e., oxides remain on the groove faces, the line will be more pronounced and the oxidation more obvious.
As for interpretation of the radiograph, the film interpretation is based on differences in film density. Darker images indicating less density, i.e., less thickness indicating a crack, incomplete joint penetration, etc. The area through the sound weld is thicker, thus it appears lighter on the film. A good radiographer should be able to discriminate between weld (lighter), base metal, darker, and unacceptable indications (typically darker).
There is no reason not to grind the weld flush if there is any concern that the radiographer will question the film due to roughness on the root side provided it passes the applicable visual acceptance criteria of the welding standard. Most welding standards (with the exception of ASME Section IX) hold the welder performance test to a higher standard than the production welds, so make sure the appropriate criteria is applied.
D17.1 is the most goofy welding standard imposed on the welding industry. Make sure you understand what it requires.
In my most humble opinion, aluminum performance tests should only be subjected to bend testing using a wrap-around bending fixture. RT will not usually disciminate between sound aluminum and aluminum oxide trapped within the weld since both have nearly the same density. Oxides don't usualy break down in the weld pool because of their higher melting temperature (above 3200 F) and they do not float out, thus a radiographically sound weld can be brittle due to oxide inclusions. The bend test will definately show you if oxides are present. Keep in mind that the guided bend test sample must be 1/8 inch or thinner. This is in contrast to most other bend tests where the standard is 3/8 inch thick. Since you are welding 0.0625 inch thick material, this is not a concern. However, the bend radius must be based on the actual thickness of the sample, the P/M/ or group number of the base metal, and classification of the filler metal. If there is a question regarding the proper bend radius, refer to AWS B2.1 or MIL-STD-1595A. If I remember correctly, D17.1 omitted referencing the appropriate standards for determining bend radius.
Best regards - Al