Restriction on coefficients =/= 0 in VAR models in R?

Question

I would like to run a VAR model with restrictions on the coefficient matrices. I know that it is common to impose some zeros, manually or by significance. In R, the vars package includes the restrict function that estimates a restricted VAR model, thanks to OLS. In the VAR.etp package, the VAR.Rest function estimates the VAR with FGLS. However, the restrictions can only be 0 on some of the parameters. I would like to know if it would be possible to set the coefficients to any real number.

The general idea is to first estimate a bivariate VAR on some variables for which we have a lot of data, and then to impose the estimated coefficients in the matrices of another VAR, bigger, but with less data. Is this mathematically possible, and do you know an R package that would allow that? If not, does it seem possible to modify the functions of an existing package?

Thanks a lot !