All good questions!
1 - I’ll give you an analogy. If there was a lottery, and you bought the most tickets, it doesn’t mean you’ll win. Having a higher hash power increases your likelihood of successfully mining a block, but finding the “nonce” (or winning lottery ticket) is still luck.
2 - Weaker miners join mining pools. This way, they are paid out a proportion of their hash power of the entire pool. For example, if my hashpower represents 1% of the pool, I will get paid out 1% of the rewards.
As a side note: The person/group/company running the mining pool charges a fee to be part of this pool and absorbs the risk of luck (i.e. will payout its members regardless of whether 1 block was solved or 10 blocks was solved from that pool). Since a block is solved every 10 minutes, and difficulty is known, a mining pool can calculate the likelihood of successfully mining a block. Since it’s all math-based, the probability of solving a block can be inferred, so the fee they charge + luck of mining additional blocks should offset the risk of under-mining blocks given the pool’s hash power.
3 - Miners are programmed to select the transactions in the mempool that provide the highest return. They are not selected manually. So a block is usually filled with transactions that have the highest TX Fee payout for their respective sizes.
Hope this helps