Except that it isn't necessarily counterproductive to mine empty blocks to get faster propagation. The propagation time depends on the number of transactions in the block, so a miner could make an estimate of the odds of a block being orphaned due to slow propagation and weigh the potential losses of an orphaned block against the certain losses of not including fee-paying transactions.
With fees still being a very small part of the block reward, the answer may not be obvious.
True, let's look at the numbers on that. According to
Blockchain.info's orphan block graph (And I qualify this by saying that I don't know how accurate this graph is) you could pretty reasonably estimate that the odds of a block being orphaned is close to 2% (144 blocks a day, around 2 or 3 are orphaned per day). So as long as transaction fees are worth at least 1/50 of the block reward there shouldn't be any significant motivation to mine empty blocks. Currently that would mean .5 BTC worth of transaction fees per block, and as you said currently the fee per blocks is generally much smaller than that.
Is my math right on this?
EDIT: I just realized that I forgot to account for the fact that sending an empty block does not completely remove the possibility of that block being orphaned. Unfortunately I don't know exactly how one would calculate the odds of a block containing transactions being orphaned vs. the odds of an empty block being orphaned.