Not really a bug as such, but something I’ve noticed is that for ALO => AL, if you use flux in your smelting process then 8.1 tonnes of ALO (6 * 1.35) gives you 10.8 tonnes of AL (4 * 2.7), which means AL is being created out of thin air. It occurs to me that this might be a problem if you are shipping ALO in and shipping out AL since the product will take up more shipping mass than the input.
The other metals/elements are a mixed bag (bold = mass gain, italic = mass loss)
For FEO => FE, 35.4 tonnes of FEO gives you 31.496 tonnes of FE (when using flux)
For CUO => CU, 20.05 tonnes of CUO gives you 26.76 tonnes of CU
For SIO => SI, 5.37 tonnes of SIO gives you 2.329 tonnes of SI
For TIO => TI, 6.32 tonnes of TIO gives you 9 tonnes of TI
For LIO => LI, 27.5 tonnes of LIO gives you 2.2 tonnes of LI
For AUO => AU 11.58 tonnes of AUO gives you 38.64 tonnes of AU
For SCR => S, 2.05 tonnes of SCR gives you 3.2 tonnes of S
For HAL => NA+CL, 6.51 tonnes of HAL gives you 5.14 tonnes of NA and CL
For LST => CA, 2.73 tonnes of LST gives you 1.54 tonnes of CA
For CLI => I, 7.26 tonnes of CLI gives you 4.93 tonnes of I
For MGS => MG, 1.73 tonnes of MGS gives you 1.62 tonnes of MG
For TAI => TA+FE, 15.88 tonnes of TAI gives you 25.524 tonnes of TA+FE
For TCO => TC+O, 9.8 tonnes of TCO gives you 12.941 tonnes of TC+O
For ZIR => ZR+SI, 9.7 tonnes of ZIR gives you 11.188 tonnes of ZR+SI
Every carbon chain (9 tonnes of C produced) is mass gaining to some degree, ranging from 3.2 tonnes of input (4 * HCP) to 7.6 tonnes of input (4 * HCP + 2 * GRN + 2 * MAI)
Mass loss in refining is clearly what one would expect in reality, and what happens for about half of the metals/elements, but the other half go the other way, with AU being the most extreme.
The obvious way to fix this would be to reduce the volume (and mass) of the products to bring them below the mass of the input. Most are currently 1m^3, but MG (0.16 m^3) and S (0.25 m^3) are not. Such a change wouldn’t effect any production lines, only shipping.