Temporal Coding - a Realistic Model of Neural Activity?

Temporal coding  does not seem to provide biologically realistic models of computations performed by neurons. However, twenty years ago Cristof Koch stated: "Individual neurons convert the incoming streams of binary pulses into analog, spatially distributed variables, the postsynaptic voltage and calcium distribution throughout the dendritic tree, soma, and axon (Koch, 1999)."
In addition many others have followed his prophecy regarding binary pulses and temporal coding  and advised  their students to search  the neural code in temporal patterns ( Dayan and  Abbott, 2001; Gerstner et al., spiking neuron models 2002; Eliasmith, and Anderson Neural Engineering- 2004,Izhikevich,  2007).
Any student that has to take courses in theoretical  neuroscience  should ask his mentor : Why would a biological "realistic" neuron convert a continuous rich  input  into digital-like uniform action potentials.
The idea of  model reduction  is very useful  in many fields as long  as the reduced model preserves essential properties of the process.
It seems that important information, the meaning (semantics) of computations is lost if the spike is considered a binary pulse.  Every spike generated by a neuron can be different. Therefore,  adding spikes (firing rate) is adding apples and oranges, similar issue if the  representation of 'neural code' uses ISI measures.
These "binary pulses" were the problem in 'cracking' the neural code.The presence of electrical patterns within spikes solves the nature of neural code,  shows that information is "read"  "written" and processed inside the cell.
Dear reader, could you please point to some papers  that show predictions, new theories or significant results  regarding, memory, behavior, object/category recognition using temporal coding in the last twenty years?
 References:

 Dorian Aur and Mandar Jog - Neuroelectrodynamics- Understanding The Brain Language , IOS Press 2010, http://dx.doi.org/10.3233/978-1-60750-473-3-i