While I applaud your desire to experiment, and in that regard, you are on the right track, but one test with no control and a single outcome is not something you should attempt to draw any conclusion from. This is just anecdotal and not necessarily a true representation.
You might have just as well flip a coin. There is nothing in your test that points to results that are anything beyond random. For all, you know if you repeated the test you would get the exact opposite result, or that 2 out three times you would get the opposite result.
You need to repeat that test many times, randomly changing the order of which page goes up first, and other factors that could influence rankings.
By the way, you don’t need to worry about the influence of backlinks as the search engines provide search operators that will allow you to filter results based on any of the four primary signal groups.
Again, I applaud your desire to experiment, but please learn a bit more about valid test methods and the tools that are available to filter results, before making conclusions and posting results that are meaningless.
Please consider repeating your test a sufficient number of times, using controls to gather useful data and share if you are willing.
When you said, “on-page SEO has so little to do with rankings I am not sure why people waste all there time on that?” This is so… not true.
Google, like other search engines, use four primary groups of signals to determine relevance. Anchor text values from backlinks is just one of those four signal groups. To suggest that the other 3 signal groups have little to do with rankings is far from true.
Yes, if you have weak competition from those other signal groups, then you can outrank someone based solely on anchor text relevancy, but it is going to be much easier if you have high relevancy from at least one or two of those other signal groups. For an example of just how hard it would be, look at the number of backlinks with “click here” anchor text that it took for Adobe to outrank most of the competition. Then, along comes a little website that actually optimized the other signal groups and knocked Adobe from that position, all without the aid of massive relevant backlinks.
While I am not actually encouraging you to disclose the website addresses, I just wanted to let you know that there are search operators that will allow you to compare the ranking of your 2 sites without the influence of backlink anchor text factors. Which in this case you should be doing anyway to isolate the on-page factors from any inbound anchor text influence.
Since it seems you are unaware of those operators and haven’t used them in your experiment, I wanted to include that as a suggestion to ensure that you are comparing the precise data that your test was designed to track.
Rather than comparing the regular search results, which include factors outside the scope of your test, you should, at the very least, be using the allintext: operator to isolate the ranking factors to just that group of signals. By using the allintext: operator, the allinurl, allintitle, and allinanchor signals will be isolated from your test. You will have a much cleaner set of signals with which to compare.
Refill, refilling, and refillable are all different words. But, more importantly, keyword density has so little impact on rankings that you will never see anything measurable except where you have virtually no competition. In those cases, it doesn’t matter much since you can easily outrank the nonexistent competition.
Instead of simply repeating the keyword, use it prominently in the content. Use it in the headline, in the first sentence of the first paragraph, make it stand out by using a larger font-size, different font-family, or italicized, bold font, underlined or contrasting color. You get far better results by enhanced fonts then you get from simply repeating the keyword hoping for some magical density level.
Anytime you try to optimize a single page for more than one keyword, you should prioritize your optimization efforts. Choose one keyword as your primary keyword, optimize for the primary keyword and then slip in the secondary keywords where it makes sense.
In many cases, it works better to simply create additional pages and optimize each page for only one of your targeted keywords.