A few tricks: try to use the center of each band, as they're pretty nonlinear. You'll notice this when you measure a +5 dB above one decade scale, and then go up one decade and see that it now reads -3, for example.
So be aware of this when making curves across decades. They'll be scallopped.
If you check the freq response of the meter you'll notice its resonant bump is pretty large. You have to calculate this peak out for a truer reference response.
Non-linearity error also increases as the battery weakens, as it requires MORE signal to push the needle up past +3dB or so.
Take special pains to use a tripod, and try horizontal and vertical orientations, noting carefully which 1/3 warbles are most effected by your hand/body/head placement.
Decide measuring distance carefully, and do NOT move either the speaker nor tripod at all if measuring comparisons.
1m distance is tricky. 2m is better for in-room response, and 2" is better for individual driver response.
Once you get good at it, you can easily measure with +/- 1/4-1/3 dB repeatability, so it can be a VERY useful tool.
Just be careful to note that it is NOT an Earthworks TC30k or B&K measurement mic!
Have fun.