Understanding mixture distributions

Yes that sounds correct. If you are observing each coin once, your likelihood should then be Bernoulli (or mixture of bernoullis)