IndexDB seems to be deliberately dumbed down in capability just to make it easier to describe in a standards document.
If we were all suffering from interoperability problems with WebSql there might be an argument for something simpler. But that's not the case -- wherever it's implemented, WebSql works the same. It turns out a good open source implementation makes a really good standard. It leaves no room for ambiguity.
And anybody who really wants to do their own implementation has free access to everything they need to know to be compatible. The fact that no one has done so just shows that sqlite3 is already good enough, and it's silly not to just use it.
When I port from WebSql to IndexDb, I will need to implement in Javascript a bunch of things that sqlite3 is already doing for me in C.
I actually started out with a very NoSQL-like architecture on top of WebSql that's closer to the IndexDb model. But when I started taking more advantage of sqlite3 to do the heavy lifting (joins, nested subqueries, etc) I got an order of magnitude better performance. Switching to IndexDb means going back to the slow way.