I think it's correct to unlink likelihood of arrest with crimes committed, and plenty of studies have shown biases in that aspect as well as the effects of coercive plea bargaining that inflate crime numbers beyond actual crimes that even happen, in addition to conviction rates of those that are arrested. However, it makes little sense to say that somehow merely having a strong incentive for avoiding arrest means that one is actually able to consistently avoid arrest. In the study, which deals with violent felonies in Texas, arguably every person who is a suspect would have just as much incentive to avoid arrest, since immigration status is not a datapoint that is collected prior to arrest and one's ability to avoid arrest is also dependent on the police's abilities and the circumstances of each case, and not entirely down to the individual ability or incentive to avoid arrest. Furthermore, since immigration consequences come after the term is served if convicted, and violent felonies tend to carry terms of incarceration that are hefty and does not differ on account of immigration status, I don't think it's really all that relevant to consider incentives in avoiding arrest as a meaningful parameter to weigh the data in either direction.
I think what's more likely is that a) because of the harsh collateral consequences which includes all of the existing collateral consequences of a conviction plus removal, in communities with large numbers of immigrants without valid status, members of the community may be less likely to call the police to report crimes to begin with. The undocumented or mixed-status community's hesitation to call the police is pretty well-documented, although that is unlikely to explain the entire difference documented here; or b) there's a self-selection bias where many of those prone to commit violent felonies may not find it worthwhile to go through the trouble of crossing and attempting to establish themselves under the radar here only to get caught up in a system that's punitive and would lead to removal regardless, considering that migrants seldom move here for reasons other than work opportunities anyway. But again, because of how byzantine our legal system is and how even more kafkaesque our immigration system is, it's unlikely that those without law degrees who specialize in these areas truly have anything close to full knowledge of what the consequences truly look at, so I don't think that adequately explain the difference either. Perhaps more studies on those aspects should be conducted, although this study is more or less in line with thinktank and advocacy group-conducted research from the likes of Cato and the ACLU, whose ideologically leanings are not exactly overlapping on most issues.