Give us some citations to document performance loss, and further citations do document that the performance loss is sufficient enough to reduce the amount of work the staff actually carries out (if they're less productive per unit time spent working, but still completes the same amount of work, then the cost is not borne by the government - it is fairly normal for workload to be quite elastic), and we can talk about your numbers.

You also don't account for another factor: Even if there is a loss, it is not given that it is a net negative: License payments to Microsoft leave Valencia, and Spain. Payments to any additional training and support staff or additional assistance for LibreOffice is likely to stay in the local economy, and part of it will find its way straight back in in the form of taxes.

It is perfectly possible that they've messed up the numbers, but you've not presented anything resembling a case for that.

You might also note from the article that this is not Valencias first large scale migration - they claim to have saved 30 million so far from switching the regions schools to Linux, so chances are they actually have better numbers than the ones you've made up.