Seeing the example searches and data would be useful indeed. Without them, though, we can still do some handwaving toward an answer. This may not be correct, depending on your exact searches, but it should get you in the ballpark.
Let's start by putting all three of your searches together. This is far from the most efficient way, but it's what works with all I have to go on:
search1 | append [ search search2 ] | append [ search search3]
The append search command will run each of these three searches and append their results together. The additional word search in the subsearches for search2/search3 is required as part of the subsearch.
Now, I'll make an assumption that "search1" returns an extracted field named "up", "search2" returns extracted fields named "failed" and "success", and "search3" returns an extracted field named "down". I'll also assume that the field names "up", "failed", "success", and "down" are unique and exclusive to their respective searches. (That is, "search1" does NOT have an extracted field named "failed", "success", or "down" ... and so on)
Given the above assumptions, we know that the stats operation max() should only find a single value to be "the" maximum for a given field. So, we can use stats to collapse this into a single table row:
... (above search) ...
| stats max(up) as up, max(failed) as failed, max(success) as success, max(down) as down
Giving you a single table row for results from all three searches ...
