I'm confused by the command line that you say you want to use:

ssh user@machine_name perl sar1.pl -e start -p -wW -i 3
[download]

Is it the case that you are really using a parameter file whose name is "-wW"? Why?

Anyway, if the script runs as intended with that command line when you run it in an ssh login shell on the remote machine, then maybe your environment via "ssh you@host command line" is different from the login shell environment that you get when you just do "ssh you@host".

You could test that by doing the following from your local host:

local$ ssh you@remote.host
remote$ env > my-remote-login.env
remote$ exit

local$ scp you@remote.host:my-remote-login.env .
local$ ssh you@remote.host env > my-remote-nonlogin.env
local$ diff my-remote-login.env my-remote-nonlogin.env
[download]

Sometimes, the login connection uses a different or extra shell script (e.g. .bashrc) compared to what a non-interactive connection uses (e.g. .bash_profile); and sometimes this could affect how PATH is set, etc.