Can this value really ever be null, or is the C# compiler just confused by code?

I've found some code that runs afoul of the new nullable reference types in C# 8, saying i could be referencing a null, when i cannot figure out how it could possibly ever be null:

https://dotnetfiddle.net/AoEMzp

public static DateTime JoinNullableDateTime(DateTime? date, DateTime? time)
{
   if ((date == null) && (time == null))
      return DateTime.MinValue;
   else if ((date != null) && (time == null))
      return date.Value.Date;
   else if ((date == null) && (time != null))
      return DateTime.MinValue.Add(time.Value.TimeOfDay);
   else
      return date.Value.Date.Add(time.Value.TimeOfDay);
}

Both Visual Studio and .NET Fiddle return the warnings for the last line:

return date.Value.Date.Add(time.Value.TimeOfDay);

.NET Fiddle

Visual Studio

So my question is, can anyone give any examples of values for date and time that would cause a NullReferenceException in this code?

At the start, the possible combinations of null-ness are:

if ((date == null) && (time == null))`
   return DateTime.MinValue;

At this point the possible combinations of null-ness are:

else if ((date != null) && (time == null))
   return date.Value.Date;

At this point the possible combinations of null-ness are:

else if ((date == null) && (time != null))
   return DateTime.MinValue.Add(time.Value.TimeOfDay);

Finally the possible combinations of null-ness are:

So at this point it is not possible for date nor time to be null. Both are guaranteed to be NOT NULL.

else
   return date.Value.Date.Add(time.Value.TimeOfDay);

I cannot see any way in which either date or time could be null on the final line - yet there it is:

CS8629 Nullable value type may be null.

So what am i missing?

In fact the nulls go away even faster

After the first line

(date == null) && (time == null)

We know that it is still possible for one of them to be null - but not both. So then we hit the 2nd check:

(date != null) && (time == null)

Before this line we knew that only one was null. And now we've found the one that it was: it was time.

This means by the 3rd line it is impossible for date to be null. And yet i added a check anyway, because the compiler told me to, and the compiler is infallible. But i have the check anyway:

(date == null) && (time != null)

So the extra checking even on the third if was redundant. And certainly it is redundant on the 4th if. And yet there is the compiler.

What am i doing wrong?

I'm certain i could re-write the function into something harder to understand, so it doesn't confuse Roslyn. But this isn't about this function, it's about the pattern being used - and if this pattern has lurking NullReferenceExceptions: i want to know about it!

Bonus

I cross-checked with .NET Fiddle in case it's an artifact in my Visual Studio.