tau decays issue

Hi,

I'm currently trying to generate events with MadGraph for the ttH, H->tautau process, taking into account the decay of the taus.
I guess that the hadronic decay of the taus cannot be directly handled by MadGraph so I only require a leptonic decay for now.
I generated the process using the following command line
generate p p > t t~ h , (t > b l+ vl), (t > b j j) , (h > ta- ta+, ta+ > vt l+ vl)
The decay of the tops seem to be compatible with an on-shell top, followed by an on-shell W decay.
However for the leptonic tau, I get a rather weird distribution, since apparently it goes through the decay of an on-shell W as well, and the invariant mass of the (vt, l+, vl) system is not at all compatible with an on-shell tau, since it is between 80 and 120 GeV.

I also tried to generate the following process
generate p p > t t~ h , (t > b l+ vl), (t > b j j) , (h > ta- ta+, ta+ > vt l+ vl $ w+)
In that case, the invariant mass distribution for the (l+, vl) system is not consistent with an on-shell W and grows linearly between 0 and 50 GeV.
Moreover, the invariant mass of the (vt, l+, vl) system is still not compatible with an on-shell tau, though different from the previous process. I get a distribution between 20 and 120 GeV with a mean of 71 GeV.

Does anyone know what would be the better way to deal with the tau decays?
I read that they could in principle be handled with TAUOLA (which should be able to deal with hadronic tau decay as well) but I didn't find an explicit example about how to do it.
Is there any user guide for that?

Thank you for your help
Cheers