How to choose the reduction function in rainbow tables?

I have a question regarding Reduction Functions in Rainbow Tables.

If the hashing function is MD5 or SHA-1 etc then should the reduction functions also be MD5 or SHA-1? That is, should the Reduction Function be the same as the Hashing Function but of course with a different set of rules/constraints for input and output parameters?

So, basically should the reduction function algorithms be same/similar as the Hashing Function algorithm but just differ in terms of rules/constraints for input and output parameters?

Let's consider that we want to recover a password $pwd$ knowing $hsh = H(pwd)$ where $H$ is a hash function. So we will compute a rainbow table for a certain set of passwords $\mathcal{P}$. Note that when the hash chains are computed, at each step $i$ we apply the hash function $H$ and a reduction function $R_i$ (note that several reduction functions are used to avoid collision problems).
A distinction must be made between the hash function $H$ and the reduction functions $R_i$. The reduction functions are used to map hash values produced by $H$ back into values in $\mathcal{P}$.
So the way to define the reduction functions $R_i$ depends on how you chose the set $\mathcal{P}$. For example $R$ can be the decomposition in base $n$ where $n$ is the size of $\mathcal{P}$ (the incrementation functions $R_i$ can be simply defined by incrementing the input of $i$ and then apply the decomposition).