A $graph$ $metric$ on a set $X$ is any function $d: E_d \to\mathbb R_+:=\{x\in\mathbb R:x>0\}$ defined on a connected graph $ E_d \subseteq[X]^2:=\{A\subseteq X:|A|=2\}$ and such that for every $\{x,y\}\in E_d$ we have $d(\{x,y\})\le\hat d(x,y):=\inf\big\{\sum_{i=1}^nd(\{x_{i-1},x_i\}):\{x,y\}=\{x_0,x_n\}\;\wedge\;\{\{x_{i-1},x_i\}:0<i\le n\}\subseteq E_d \big\}$. A graph metric $d$ is called a $full$ $metric$ on $X$ if $ E_d =[X]^2$. A graph metric $d: E_d \to\bar{\mathbb R}_+$ is $floppy$ if $\hat d(x,y)>\check d(x,y:= \sup\{d(\{a,b\})-\hat d(a,u)-\hat d(b,y):\{a,b\}\in E_d \}$ for every $x,y\in X$ with $\{x,y\}\notin E_d $. We prove that for every floppy graph metric $d: E_d \to\mathbb R_+$ on a set $X$, every points $x,y\in X$ with $\{x,y\}\notin E_d $, and every real number $r$ with $\frac 13\check d(x,y)+\frac23\hat d(x,y)\le r<\hat d(x,y)$ the function $d\cup\{\langle\{x,y\},r\rangle\}$ is a floppy graph metric. This implies that for every floppy graph metric $d: E_d \to\mathbb R_+$ with countable set $[X]^2\setminus E_d $ and for every indexed family $(F_e)_{e\in[X]^2\setminus E_d }$ of dense subsets of $\mathbb R_+$, there exists an injective function $r\in\prod_{e\in[X]^2\setminus E_d}F_e$ such that $d\cup r$ is a full metric. Also, we prove that the latter result does not extend to partial metrics defined on uncountable sets.