If there were no rewards, than most of miners would quit so those that would stay would make profit (self regulation by market). Thing is, as mining rewards go down, btc expansion should go up (one not connected to another). With expansion come more transactions, with transactions come more fees, so basicly, in far future, fees should be able to replace block rewards. Not maybe 1:1 (they can be even bigger as far as theory goes) but they would be profitable for miners that stay and mine after block rewards drop down.

Assuming miners are rational (which they're not), then there will always be "just enough" miners to keep the difficulty at a profitable level - even when there is no block reward.

They are reporting the deficit as txn fees (not including the block subsidy of 25 BTC) minus their guesstimate of what miners are paying using unknown method of computing that.  Since blockchain info provides no context or supporting details for their estimate of miner cost I would treat it as worthless.  

For about a year after the release of ASICS they used the average GPU efficiency and an electrical cost of $0.15 per kWh for similar charts.  This lead to an underestimating of effieincy by at least a factor of 100x and lead to a more than one false news stories about how bitcoin uses $50M a day in electricity (or some equally ridiculous number).  This is despite countless request from lots of people for them to fix the "stat" or just delete chart (no info is better than wrong info).  They didn't for more than a year so accuracy of stats is not high of their list of thing to do.

Not that it matters for the reasons above but deficit (or surplus) already implies revenue - cost.  So if the deficit is -$2.4M then blockchain.info is saying miners would be losing $2.4M not $11K a day by mining if the block subsidy was zero.