The core reason why the electoral system is so sensitive is the winner-takes-it-all allocation of electors. Just a small change in total votes (e.g. as a result of redrawn borders) can swing all Florida’s 28 electors to another party.
This sensitivity issue could, of course, be tackled by using the popular vote. However, there are several (good or bad) reasons why many people think that places such as Vermont and Idaho should be over-represented and places like California and Texas should be under-represented in federal elections.
An alternative method would be to allocate electors proportionally in each state. For example, if we would allocate California’s 55 electors proportionally (by using D’Hondt method) Democrats would have got 35 electors, GOP 18 electors, and both Green and Libertarian parties would have got one elector each. As you see proportional election would be less sensitive to gerrymandering or the special interests of small swing groups. In proportional election if you gain or lose 10,000 votes in Florida, the swing would be one elector, not 28 electors. It also makes the election more meaningful for a red voter in a blue state and vice versa when there is a real possibility that your vote counts. Indeed, with this method both main parties would have got at least one elector in each state except in D.C. and Wyoming.
So, if we would allocate electors proportionally at state level (and by keeping the number of electors per state same as now), the electoral college would look very interesting:
Clinton 267 electors
Trump 267 electors
Johnson (Libertarian) 2 electors (CA, TX)
Stein (Green) 1 electors (CA)
McMullin (Independent) 1 elector (UT)