Hello Author,
The reported problem is registered now in OCCT Mantis bug tracker under ID #29085, see https://tracker.dev.opencascade.org/view.php?id=29085
Should you find other cases of failure of fillets or other algorithms, please report them to OCCT Mantis directly. Note that preferred way for reporting the issues is OCCT DRAW script, as it allows developers to reproduce the problem easily. It is also OK to provide C++ code, while this assumes more efforts needed to reproduce the issue (hence less chances to get triaged soon). You can attach shapes (in BREP format) or data files in formats supported by OCCT (e.g. STEP or IGES files). Please be specific in description of the problem and provide all details to understand it and to know what is expected (as you did in this forum topic). Images illustrating the problem are always welcome; they can be attached to the issue.
Commenter-1