For a "glitchless" run, the time is about 55 seconds higher than it should be if it were optimized. Even for a real-time speedrun this would be a fairly poor attempt; there are way too many pauses or slowdowns everywhere, as well as sloppiness such as running into walls, which kills your momentum in the process and costs multiple seconds at times. Other things that stood out:

• In 1-1, not only are you already 4 in-game seconds slower than the fastest real-time runs and TASes, but you also get fireworks, which cost even more time.
• In 4-2, you decide to hit a single invisible block which is too high up to reach immediately, so you just end up waiting for the moving platforms on the left - twice - instead of just hitting some other blocks and climbing up that way. This costs a massive amount of time. Additionally, at the end, you undershoot the point where you need to turn left twice, so you end up dancing on the ledge for a second. You have rerecords to remove these sort of mistakes (along with many others), why didn't you do this?
• In 8-1, you stomp a Paratroopa from below and clip through it, which can be considered a glitch.
• In 8-2, before the large gap, you end up dropping your speed sufficiently that you need to turn back on the pipe in order to take a running start and jump across the pit. All of this could have been avoided with more careful speed management. Also, after score countdown, you mash buttons a lot (which you do in general during intermissions), which in this instance causes you to pause the game, needlessly wasting a few more seconds.
• In 8-3, you do a "perfect" jump over a piranha plant multiple times, clipping through them in the process. This is the number one reason real-time records of Super Mario Bros. are accused of cheating, and a "glitchless" run should avoid such instances if it wants to remain true to the spirit of its category.
• In 8-4, before entering the midair pipe, you miss the invisible block before it twice, which really just looks like an epic fail. This would be absolutely terrible for a real-time run, but a TAS just has no excuse, period. You also clip through a fireball and Bowser's hammers at the end, further making it seem less legit for a glitchless run.

Besides this, "glitchless" is not a good category for this game, as the few glitches present do not make a large difference timewise in this game. I'd expect a fully optimized glitchless runs to be maybe 10 seconds at best longer than the "glitched" run, and even then it's still not publishable as it is very redundant to the main run. Also, Super Mario Bros. is actually not a good game for a new TASer to try and run, as the existing runs are already very thoroughly optimized and very hard to improve, and they also set a very high standard for runs of the game. Bottom line: 74 rerecords are definitely far from enough for a run of this length; I'd expect at least two orders of magnitude more rerecords for a run of this game. The 74 rerecords in this run don't even seem to have been used all that well, due to all the pausing and sloppy errors made in this run, which aren't even excusable for a real time speedrun without rerecords. kusoman, I recommend you take a good look at runs like [1715] NES Super Mario Bros. "warps" by HappyLee in 04:57.31 to see what kind of quality we expect for a tool-assisted speedrun to be accepted on this site. Be sure to also read and understand the Wiki: Guidelines in order to see what is needed for a TAS worthy to publish on the site.