The challenge of assessing the psychosocial working environment: why some self-reports should not be interpreted as environmental exposures