Why do you expect the genomes to have the exact same sizes? "Closest" sequenced homolog can still be far far away. I would say you go for more meaningful assembly quality assessment criteria.
Alternatively, you could just randomly try assemblers until you find one that gives you ~20MB
