But is the spectral sensitivity of cheap meters really same as in the definition of the photometry units ?
Within their accuracy specification...
CdS matches the vision characteristic pretty well, but the general sensitivity tends to long term drift over time.
Semiconductor sensors are worse, mainly because you don't exactly know what the sensitivity shape really is.
Of course, the best is a calibrated spectrum analyzer and then an (Excell or so) math, but that is more difficult to get, although many DYI designs are around - mainly the ones using CD as difraction grating and then some camera, all the setup then relatively calibrated using a regular incandescent lamp (the incandescent provides well defined accurate reference spectrum, the only thing you need to know to put as a parameter is it's temperature)