We (I) get asked questions like this all the time, ie. round up or down, or how much margin of error.

I can't advise because I have no way of knowing the original bias in spoke length. My personal guideline is +2,-3mm from the top of the nipple (for double wall rims) or +0, -3 for single wall rims, unless you want to grind afterward. Note, that I'm careful to use nipples that allow 2-3mm of overrun, otherwise, I can't err high.

Since you have wheels with known data, calculate those, and compare the results to the actual spoke length, and actual position with the nipple to better estimate the bias of the calculator, then calculate the wheel you want to build, and apply an appropriate fudge factor and see if you can have all spokes within tolerance using only one length.

For the record, it's generally possible to use a single length spoke, with it being acceptably longish on half the wheel, and acceptably short on the other half. But that narrows the working tolerance, so using two lengths is preferable because it reduces the odds that you'll have to built twice.

Lastly, you need to maintain perspective. Having spokes end well into the nipple head is stronger, and prevents nipple breakage. But there's room there, and lots of wheels are built with spokes that I'd consider to be too short. It's not like these only last 10 minutes, so if this isn't a bike that you expect to put thousands of hard miles, and especially since it's a front wheel, I wouldn't get overly worked up about spokes being a bit short.