The problem is the proof of work is tied to the transaction that sends the block reward.  So, there isn't an obvious way to generate a coherent block using multiple proof of work solutions.  So, the ways to make the block reward more granular means more blocks more often.  However, there is a tradeoff as each block has additional overhead.  So if you change from 10minutes to every 1 second the amount of additional network / computation / storage would make running the protocol much more challenging.   So, there is the trade off.