Initially the annealing temperature should be at (or slightly above) the Tm of the complimentary sequence between the template DNA and the primers (i.e. excluding the 5' flanking region). If you have a gradient block PCR machine I would set up a run with annealing temperatures ranging from 60-70 deg C across the block and pick the highest temperature for which you get a single band.
Keep the Mg 2+ to a minimum as the amplicon is small so you can sacrifice efficiency for specificity and ensure the elongation step is around 15-20 sec (check the elongation rate of your enzyme and have the elongation step slightly under the calculated time for your lenght of amplicon).
If you still obtain nonspecific bands you might want to try the initial cycle at or above the Tm of the complimentary parts of the primers followed by an increase to 65-68 deg C (for the annealing step).
I tried the annealing temperature for given set of primers excluding the 5'flanking region (before trying a gradient PCR, as I was too much excited to give it a go) and it worked.
the annealing temperature is calculated only for primer sequence, that complement to target sequence. Additional tail at 5'-end of primers is not important for Ta calculation.
I want to give the same comment as given by Ruslan Kalendar. For primer with flanking regions, Ta is calculated for only specific primer that will being to DNA sequence.
I have a different answer. I used to calculate the Tm of the primers with the flanking 5' tail regions in both forward and reverse primers. This is important because, in the first PCR cycle only it will act as a tail and in the subsequent cycles it will be a part of thee amplicon getting amplified. Even in very rare cases I excluded the tail region and calculated the Tm and got success, but I prefer to calculate Tm of primers with the tails. Its depends on individuals and through experience, one will decide to add or not to add while calculating Tm.