Can someone help me on response surface optimisation by state-ease design expert software for compositional responses [% composition of responses]

I want to decompose a particular type of biomass in a reactor under controlled conditions of temperature and residence time. I want to optimise the operating conditions of the reactor by varying the temperature from 200 to 500 degrees Celsius and residence time from 5 to 50 minutes. The responses are four decomposition parameters of biomass, which must sum up to 100%. Let say when heating the biomass at a certain temperature and residence time, the biomass decomposes to four products [X%, Y%, Z%, M%], sum of which should be 100%. So I want to optimise one of the product, let say X%.

I did the response surface methodology [two factors] and conducted the required number of experiments.

In the analysis, I can see the expected behaviour of responses with the variation of the two factors. But the problem comes on the optimisation side, the optimum conditions suggested by the software gives the responses that exceed 100% , which is not feasible.

Can you please comment on it, like what design I was supposed to use or what is the optimum responses suggested tells about.

Thank you