[Bug] Multiple judgements with mental operation may derive wired conclusion

[biaktfn]

Describe the bug

For example, we can input:

1: (^want,{SELF},<a --> b>). %1.00;0.90%

And it will create a goal:

2: <a-->b>! %1.000;0.900%

Then we can also ask:

3: <a --> b>?

It will create a judgement:

4: <(*, {SELF}, <a-->b>)-->^wonder>. :|: %1.000;0.900%

The inference rule can apply on 1 and 4, so there will be:

<^wonder-->^want>. %1.000;0.299%
<^want-->^wonder>. %1.000;0.299%

which are obviously should not be derived.