Unexpected performance metrics for "sorted_arrays_merge" in C++

[ArshaShiri]

Hello and thank you for this great resource.
I implemented a simple brute-force solution for "sorted_arrays_merge" problem in C++ as below:

vector<int> MergeSortedArrays(const vector<vector<int>> &sorted_arrays) 
{
  auto result = std::vector<int>{};

  for (const auto &sortedArray : sorted_arrays)
    result.insert(result.end(), sortedArray.begin(), sortedArray.end());

  std::sort(result.begin(), result.end());
  return result;
}

I get the following metrics on my machine for this solution and use the default compiler flags:

Test PASSED (152/152) [   9 ms]
Average running time:  153 us
Median running time:    44 us

If I run the provided solution, I get these metrics:

Test PASSED (152/152) [  19 ms]
Average running time:  377 us
Median running time:   155 us

I tried to improve the runtime of the provided solution by using a class like this to avoid iterator indirections:

struct ArrayInfo
{
  bool operator>(const ArrayInfo &other) const { return el > other.el; }

  const vector<int> *array;
  size_t arraySize{ 0 };
  int arrayIdx{ 0 };
  int el{ 0 };
};

However, that did not change the duration that much. I was wondering what the issue could be that the optimized algorithm runs slower than the brute-force one. Thank you very much for your time and help.