Hi Urs,

sorry for this short response but I have to leave in 5 min....
So concerning the 1/mb3 errors - these errors are supposed to cover
the fact that we use 1/mb3 expansions for mb(lmabda) and l1
but ignor the other 1/mb3 operators (e.g fix them to zero).
Therefore, the extracted values Lmabda(1/mb3) and l1(1/mb3) have errors
that stem from the above mentioned procedure. When you break them down by
one order e.g. Lambda(1/mb3) -> Lambda(1/mb2) one has to include the 1/mb3
uncertainties also in Lambda(1/mb2) because the starting point was the
extraction at O(1/mb3) - right?

Therefore, the used set of Lambda(1/mb2) and l1(1/mb2) obtained from the
1/mb3 CLEO OPE interpretation should also have (1/mb3) uncertanties - in
fact they do, because the ~100 MeV theory error on Lambda(1/mb3) get
propagated into your Vub theory error - right?

On Wed, 24 Sep 2003, Urs Langenegger wrote:

>
> Hoi Oliver,
>
>  > Another important issue Ed completely ignores , is the fact, that the OPE
>  > parameters Lambda and l1 have rather large uncertainties due to the
>  > unknown 1/mb3 corrections.
>
> I am  not sure about  this point any  more.  As all this  (SF relation
> between b->sg and b->ulnu) is  leading order in mB only, we explicitly
> remove higher-order terms to be "as consistent as possible" (though we
> are not  really, we need the 1/m^2  terms, not only the  1/m terms, as
> lone only  comes in at 1/m^2).   But this might  imply that subleading
> corrections could  be expected to contribute  and then we  are back at
> the higher corrections...
>
> For all  the rest we agree, I  think.  FWIW, I believe  that the error
> quoted  by  Ed is  only  the statistical,  and  does  not include  the
> systematics (which are smaller for the measured <E_g>.
>
> Cheers,
> --U.
>
>
>