1) Estimate of uncertainty of BR(t -> H+b)

We assume Ga_tot = Ga(t -> H+b) + Ga(t -> W+b) with CKM = 1.
The latter assumption relieves us from taking into account lighter quark
decay modes.
delta Ga(H+b) we have from the Delta_b variation (as discussed in the
last meeting).
delta Ga(W+b): we include the QCD corrections here, so missing are 1L EW
and 2L QCD. I think that assuming 5% for 1L EW and 2% for 2L QCD is
conservative enough.
The variation in BR(H+b) can then be obtained from

delta BR(H+b) = [Ga(H+b) +- delta Ga(H+b)]/
                [Ga(H+b) +- delta Ga(H+b) + Ga(W+b) -+ delta Ga(W+b)]

The last '-+' is chosed to 'maximize the uncertainty effects'.
Theory errors are added linearly.



2) Estimates for Ga(H+ -> tau nu) and BR(H+ -> tau nu)

Let's assign the following (conservative) theory uncertainties:
delta Ga(tau nu) <= 5% (missing 1L EW)
delta Ga(AW)     <= 20% (missing top/stop loops)
delta Ga(HW)     <= 20% (missing top/stop loops)
delta Ga(t*b)    <= 10% (possibly missing higher-order QCD, ...)
delta Ga(cs)     <= 30% (see the unresolved issue from Markus' analysis)

delta BR(tau nu) can be evaluated from the above formula.

In the mhmax scenario, which we are most interested in, only Ga(tau nu)
and Ga(cs) play a relevant role, all other channels are far below 1%.
This yields a conservative estimate of <= 1% uncertainty in BR(tau nu),
which, according to Markus' analysis, can be evaluated with FH or FH/HD.