From a statistical standpoint, employment in the U.S. is broken down according to the North American Industrial Classification System (NAICS) category of the employer, not the job. So all employees of General Motors, for example, were classified as "manufacturing" employees even if they had absolutely nothing to do with the production of automobiles or any associated components.
This means that an accountant who worked for General Motors was considered a "manufacturing" employee, but if he were laid off by GM and hired by an accounting firm he was considered a "professional service" employee -- even if he spent 100% of his time at the accounting firm doing accounting work for GM.
The biggest influence in the so-called "loss of manufacturing jobs" over the years has not been outsourcing, automation, or anything of that sort . . . it has simply been the re-classification of employees into their proper job classifications as these major manufacturers have shed a lot of their back-office functions and hired outside firms to do that work.
You've just mentioned one reason why I don't get intimidated by statistics. Heaven only knows what some of these numbers actually represent.
People on workfair working part time and making minimum wage are considered to be employed. People who are attending training courses paid for by the government are sometimes considered to be employed.
There are times when numbers can be compelling, but the days where they could be considered to be above manipulation are over IMO.