Works better with "source_anime, detailed, 1970s \\(style\\)), vintage" in prompt. Use of "flat shading, flat color" seems to help if you aren't going for a headshot only (the show would use a bit of shading in facial closeups to make faces "rounder" but not bother in body shots), especially on the face to make it a bit rounder). Even then though, it seems very picky about other resources it will work with: Even character loras otherwise good and accepting styles will just seem to ignore the Gundam style LoRA entirely and generate an image with no style influence from this at all.

I'm curious what the training data included, because the original anime is a bit infamous for total inconsistency in animation quality.

It's a shame this never got an updated description and showcase pics, because it works really well when you figure it out.