Whenever we do this reaction using an alkyl aldehyde, we see a small amount of a side product that elutes more quickly than the main product. It gives a positive reaction to permanganate stain. IIRC when I took an NMR one time, I saw signals in the vinyl region, but I did not analyze the signals at the time (I may do so in the near future). I do not recall seeing a side produce when an aromatic aldehyde is used. The base in our Horner Wadsworth Emmons reactions is typically DBU. I imagine that our aldehyde is somehow doing an elimination, but the oxidation state must be accounted for, and I don't have a good mechanism thought out. As long as this is a minor reaction, I doubt that it is costing us much in yield, but I want to think ahead in case it does in the future.