I've recently assembled a plant genome. It looks fine, with a scaffold N50 of 850K, and scaffolds longer than 10K take 90% of the total length. But here comes the question: the assembled genome size is a little larger (230Mb)than kmergenie prediction (205Mb). I've checked the histogram fit of kmergenie, and it looks quite good.
In my experience, the result of Kmergenie is usually smaller than the old method (kmer_num/depth_peak) and the assembly size. Does anyone know which number is more accurate?