Well, sure, objectivity is the most practical approach, but that doesn't make it the most representative of truth. Similarly, solipsism isn't useful, but is logically sound—or we usually talk about the past and future as actually existing when we can't prove that they are. What this then comes down to is which we value more highly, utility or truth.

Regarding the senses being used to error check the senses, at best we can say that the later sense perception seems more reasonable than the earlier because it is more consistent with what we believe we have previously perceived. The determination is based only on consistency.

If you were to write a piece of code that worked sometimes and not others, you would have to admit that the code is faulty or should be rewritten. You are arguing for writing a second piece of code for finding consistency in the outputs of the first piece of code and then throwing out inconsistent data from the final output. Unless the second chunk is very good you run the risk of undue biases, for example, confirmation bias. For example, both the theist and atheist will remember the experiences that confirm their core belief and forget those which challenge it. The adept programmer would rewrite the first function from scratch to reduce fault points. 

Beyond that, there are other means for testing for truth, for examples reason & introspection. Of course, the kicker is that without having boot loaded ourselves based on sense perception experiences early in life, both reason and introspection would have nothing to think about other than the '<i>I</i>'.