As mentioned there about 6 days ago:
is the pts code to monitor ram usage wrong and there is still no fix in git. Also you never commented on that bug, why not? Do you think it is correct? It is not logical to take buffers + cache into account, even free has got a "-/+ buffers/cache" line which could be used, but definitely not what you use now. Whenever you have got disk access - while installing tests or whatever buffer (or cache) will increase, simple code to increase buffers (any code can be used, just this shows nice thruput values too, also nice for cloning hds):
When you monitor with a correct tool like htop the used ram will not change during run, but your tool will till it reaches a max value for buffers/cached. Just calc FREE or USED correctly and define the other just as USED=TOTAL-FREE or FREE=TOTAL-USED. I think i gave you enough examples to calculate it correctly.
buffer -i /dev/sda -o /dev/null -z512k
After you fixed the code you could retest and update your ram usage article.