The lack of microfoundations is important in second order. Lets not forget that it is, at the end, a general equilibrium model. Moreover, the IS-LM model is a synthesis of the goodnes of general equilibrium economics and comparative advantages of Keynesian economics.
Although Paul Krugman is lately a maverick economist, opposing to every other collegue, it is worth taking a look at Krugman's thoughts of the IS-LM model. Personally, I have more respect now for the model. Not because its limitations are minor, but because its usefulness has been proved over years of economic policy.
The fact that international agencies such as the IMF or central banks use the model -along with some dynamic stochastic general equilibrium- shows its value for the profession. I find it hard to believe that after tons of macro papers written out there, only a small proportion has proven useful for macro policy.
Two facts have changed my view of the model: First, I prefer to be an "engineer macroeconomist" in Mankiw's definition. Second, when learning advanced models of economic growth and policy, the IS-LM provides the right intuition. So it makes it easier to understand the important results.
Hopefully the current economic crisis will trigger interest in making economic modelling closer to reality. It is true, economic science evolves slower than it looks. Maybe modern macro models are over a path in o a new and more useful model than the IS-LM. In the mean time, it is not a waste of time to learn it well if you like macro.