Assuming the extract has some antibacterial activity, the problem may be poor solubility of the antibacterial compound(s). They can't diffuse if they aren't dissolved. A possible solution would be to add the extract to a top agar layer poured onto the plate so the antibacterial substance is already spread throughout. Each plate could have a different concentration of the extract to find the MIC. Another approach would be to do the experiment in liquid culture with shaking.
Adding to Adam B Shapiro answer, it could be a problem with the uniformity of the agar, having claps that hamper diffusion. Try pouring the agar when it's hotter.
Based on my experience with the well diffusion assay using jade plant extract, I found it challenging to accurately pour the solution into the exact well, leading to potential overflow and ambiguous inhibition zones. Additionally, variations in bacterial preparation, particularly due to overnight incubation, can result in inconsistent or imprecise measurements of the antibacterial clear zone upon repetition. This issue can be mitigated by reducing the turbidity of the overnight incubated bacteria to achieve a standard control, such as 0.5 McFarland standard. Furthermore, if some active compounds in the extract are not perfectly soluble in water, it can lead to inconsistent diffusion through the agar medium. In my view, the disk diffusion assay is preferable to the well diffusion assay. Soaking blank discs with the extract ensures the desired concentration is achieved and prevents overflow and inconsistent diffusion through the medium.