I have built Gilbert cell multipliers and they worked fine with transistors that were not matched pairs. I wonder if you got something wrong in the design. A common error is thinking that transistors are current controlled. Really it is the emitter base voltage that makes the emitter current flow. Most of that comes out the collector. The base current is the fraction that the HFE controls. On transistors, the HFE is a big number so the fraction that goes out the base is small even though it differs a lot. On a given transistor type, the emitter current for a given emitter voltage varies a lot less than the HFE does
Well the design I used is available @1:19 so if you have any suggestions for improvement I'm all ears. To answer the question about matching. Yes they do have to be matched. The reference I used can be found on the following link: oarklibrary.com/viewers/pdf/viewer.html?file=https%3A%2F%2Fapp.oarklibrary.com%2Fproxy%2F%3Furl%3Dhttps%253A%252F%252Fcurriculumoark.blob.core.windows.net%252Foark-library-container%252Fexternal-search-provider%252F4a514ab0-dbff-49f1-87c5-0514a285945e%252Fe5dd01a0-096f-401a-b1d7-0d9a8615073b%252Ff395d6a6-2173-4155-aafa-440ba5610be6.pdf If you follow their math on slides 12 and 13 you can see that they cancel out the reverse saturation current which is only possible if the transistors are matched. Perhaps you used some kind of compensation circuit in order to eliminate the mismatch problems.
